35 Burst results for "Braden"
"braden" Discussed on VUX World
"The Voiceless NLU performs 1% better in classification tasks against standard benchmarks than GBT4. GBT4 was three orders of magnitude more expensive to run. Again costs will come down over time. I think NLUs for like I mean I think over time LLMs will become better for classification than NLUs but at the present moment I think NLUs are typically faster and cheaper by such a large degree that it makes more sense to use them however I think as you know assuming sort of like NLUs reaching the you know the top of its maturity curve and I don't think it is because I think LLMs are actually going to help NLUs a lot but then I do think sort of the progress for LLMs is you know we're just at the beginning of it for the most part. It will become better at classification over time and I think what you might then see is using an LM for classification and LM for generation and then you might use maybe use a once again a chain of thought to actually you know check on the actual response itself like you know you might use an LM end to end with multiple different calls. I don't think I'm in a strong enough position to say it but I know that right now if you're going to go deploy a bot at scale NLUs are far cheaper and generally is good at classification and so it really comes down to cost and things like that. But that will change. That will change. Oh yeah 100%. I think it will all change. I think what's really interesting is how you know we're not necessarily right at the very beginning of this but we're certainly at the beginning of the technology being utilized properly within tooling platforms like Voiceflow and also definitely at the beginning of the kind of populations general awareness of this stuff. I told the anecdote before about my mom. My mom texts me. Gets a text. Are you watching question time? I said no I'm not watching question time. Why? They're talking about this computer that's going to kill us all. I'm like are you talking about chat GPT? She said well you've heard of it. I was like I fucking heard of it. Mom this is all I do every day. So it's like I know when something has reached maximum exposure and that's when I get a text from my mom saying that this stuff is happening. So it's definitely early days as far as I think seeing how we can apply this because now that people who weren't part of the Alexa movement, weren't part of the Facebook Messenger movement, weren't part of the initial kind of IVR automation sort of space. Now there's a brand new breed of people who are now finally waking up to conversational AI. They don't think of it as conversation design. They think of it as prompt engineering. They don't think of them building AI assistants. They think of building agents. But we're all talking about the exact same stuff and so what's interesting is where it's gone ahead over the next kind of five years now that we've seemed to have kind of you know the community and the technology has finally reached itself out of the power pit and it's out there in the world now you know. Totally. It's been a marketers nightmare. The amount of different terms that Voiceless has to use like we're now we're going in on the agents thing but like assistants sort of the past two years before that was skills. It's you know bots was before that. It's been as an industry let's all vow to choose one and stick with it because it makes it difficult to like you know build up SEO and know what they call ourselves too. So yeah it's been exciting. I mean I think the next five years I'm just I feel super fortunate that we're in an industry that I feel super excited to work on every day like it's been so cool seeing what the community is building and seeing everyone's super jacked up and also you know we have like the folks who've been in the community for a while now for four or five years you know you you and I have it's my third time on the podcast it's cool to see like this fresh injection of excitement where it's the biggest wave that any of us have been a part of and I think more than anything else the wave has a lot of a lot of ground to stand on. It's you know there's there's real business objectives like business progress is being met by using these technologies and I think that's the most exciting thing since you know I think in the past we've had you know I think it's sort of about Alexa in particular Alexa had a lot of momentum I still think there's a really bright future for Alexa and you know I would be surprised if Amazon doesn't make big moves in the coming years given all the LLM stuff but the initial version of the developer ecosystem had a lot of excitement and then you know it didn't hit the business value and it feels like large language models are now starting to hit that point where people are going to be able to build their careers off of it and that's going to make the industry you know a lot bigger so yeah it's going to be exciting times and you know we're happy to be a part of the community.
"braden" Discussed on VUX World
"Definitely definitely chain of thought prompting with self consistency was what I was looking for earlier on when I was talking about monitoring multiple models seeing what's consistent and then using that to be the answer but then what you need to do there is you need to have like multiple models and so doing that at scale it's expensive enough to use one model but you're using four models and then you're going to grade the responses it's going to get very expensive. Quick funny story here we I like to build like voice flow agents on the weekends and stuff like it's one of my hobbies and I was building this agent where I essentially wanted to like have like a every single time you would give it a new piece of information it would like check in its transcript it would update the previous histories and like sort of have this constant recurring memory and then as you came up against the context and token limit it would then actually look at its memory and see where it could condense like sort of remove some fidelity and over time you sort of have this like you know massive long term memory and I built this you know and I have the voice flow credit card so it had you know unlimited number of tokens to play with and then I shared the template with the community and I forgot to disclaim that hey like this might burn a million tokens a call and then I had a whole bunch of voice flow users complaining to me they were like what the heck like you like burned up all of our tokens oh god all these like really intensive cool stuff that you can do with LLM's it's very token intensive like it's like you know it might end up being like a dollar a call to do like you know some of this like chain of thought and all that kind of stuff. I did read about Microsoft potentially working on a chip which is intended to try and bring the cost of this kind of stuff down how far away or close that is who knows but yeah it's interesting. I think the only like one technology thing I think right now they're too expensive but I think as technologists we should always assume that things will get exponentially cheaper over time like that's one assumption we're making at Voiceflow is like we used to be worried about the latency when we were building our initial LLM features we thought that was going to come down it has and then we're also we were worried about cost and that came down by one order of magnitude and we think it's going to come down for more and so I think most people should be under that assumption too that like at some point LLM's will become as maybe not but like they'll be comparable to like an NLU. Yeah definitely I think it's the chips cost at the minute I think that's the time for the battle is that you know they require big computing power to run the inference and stuff like that and as long as it costs a lot of money to do that then that's kind of where it's going to come from but I think the scale the economies of scale obviously that will come with that and if there is new chips coming down the pipeline from Microsoft or whatever then inevitably it's going to get a lot cheaper a lot quicker. There was one final question I had I know we're a little bit over so apologies for rooting about. I had one last question because I spoke to Mikhail Birstev last week who is the founder of Deep Pavlov and one of the things he mentioned is that they've been doing a lot of experimentation with large language models and when they've benchmarked large language models to other NLP models the classification wasn't as effective and entity extraction wasn't as effective but I know of quite a number of vendors who've told me personally that classification is an absolute kind of like you know dream use case for large language models and so is entity extraction so I'm kind of hearing two things from I mean I don't doubt Mikhail he's a very experienced very sort of like you know shrewd sort of researcher who's got a lot of experience in the in this kind of field but at the same time people that I trust tell me otherwise so I'm curious about what your kind of opinions are specifically not the generation we've discussed generation we've discussed potential hallucinations but specifically on the classification and entity extraction side of things. so I'm not as deep on the technical side so I'll just give what I see.
"braden" Discussed on VUX World
"Nice. And you alluded to something there which is interesting in terms of one of the concerns that businesses have around large language models is the whole hallucination problem. I know a lot of big enterprises who are experimenting with this kind of stuff internally for internal you know knowledge search and stuff like that and I've seen a couple of examples like loop insurance and stuff like that where they're using it on the front end and on that lot and in that instance actually it worked pretty well but there's definitely kind of for certain types of companies a bit of apprehension and I suppose some of that is because for me there's different levels of use cases we've got this like data out framework which is like that goes from deflecting use cases which I view as the FAQ stuff which is a nice playground right now for large language models answering questions. It then goes to interpreting use cases where you need to basically it's like a one way data retrieval from a line of business system and then it starts to become a little bit more deterministic because you need to get information from someone to validate it to then retrieve the right kind of data then it moves to transacting which is two way integration so then it becomes even more deterministic because you need to get the right data in and to get the right data out and so then you start following a defined business process then it moves on to assisting and transforming and the assisting and transforming is about being more proactive using data in a more sensible way better natural language understanding which I think LLM's can definitely help with and then transforming is use cases that you couldn't have imagined or delivered previously with humans alone and so in the transacting bit the T this is where businesses are beginning to get to they've always focused on the content the search and large language models are proving that minus the odd hallucination that they can do something in that space when it comes to the next stage of maturity you start to then you need to get a bit more deterministic because business processes and policies at different companies are different so I'm curious about your thoughts on because you mentioned there around the conversation design being less about being deterministic less about what do I say here and do this for that flow and more about abstracting that yet a large language model isn't necessarily familiar with XYZ banks specific policies so I'm just curious about how you're thinking about moving from the sort of like the deflecting based use cases the FAQs and the question answered the knowledge based stuff more towards more kind of like utility based mission critical transactional kind of use cases and the role large language models can play there are the challenges of doing that yeah I think generally there's sort of two levels of assistant complexity I would actually I'd argue there's three so one is like it's just FAQ automation right essentially knowledge automation single turns like single turn conversations right then you move into like what I almost describe as like routing routing conversations where the goal is not to do end to end automation it's to get the person to the right queue for the live handoff with all the right information right so you've you know these sort of build up and then the last is that end to end automation right and so there's sort of these three different levels of of agent I think you know touching on sort of the LLMs and this stuff business logic in my opinion should still be fairly deterministic because like logic is like like what the reason LLMs hallucinate is ultimately what these models are doing is predicting what to say based off what was said previously and it doesn't necessarily have an innate understanding of your business per se you kind of have to code that in it's the same thing with Starbucks like you know if I was to hire a barista and say hey you know take this order there needs to be an innate understanding that in order to take the order I need their name what they want and how they're going to pay these are like the non-negotiables of my business right like I just you know you can I almost think about it like there's this like rigid core of business logic and instead of the the logic being rigid and the conversation used to be rigid you now have this incredibly fluid interface on top which is awesome but it's sort of gliding over top of a rigid interface of logic and I think for most businesses that like that's not going to change right I don't think Starbucks is going to say hey just because we have a large language model now it means you know they don't have to tell how they're going to pay that still has to be there and ultimately too there's a lot there's there's human accountability at what point who's accountable for the bot and that's where I think that that rigid communication comes in so I think what you'll probably see is something like this sort of using these two different these two different ideas you'll see a lot of FAQ FAQs using generative AI you'll see like large language models paired with knowledge bases I think like that level one of agent it's such a good use case because you know it handles the long tail really well it's able to synthesize information from you know different data sources that's great I think when you then go up to the agent assist I think you'll then have like exit criteria essentially of like you'll try to have end to end automation of course but you know you're not always going to have 100% containment how do you use the LLM to essentially do information retrieval but you're using rigid checks on what do we need to actually progress to the next stage of the conversation what do I need in order to progress and hand this off to an agent and then the same thing for that sort of like fully automated conversation what do I need in order to complete this conversation what's the good exit criteria what's the LLM working towards and then you're doing essentially like the rigid business logic checks to ensure that like the LLM is not like going all over the place and things like that that's kind of that's kind of how I think about it and it doesn't touch on like you know there's hallucinations for sure in that like level one I think what you'd probably want to do is you know a I think things will get better over time I think you could start to see some interesting paradigms like an NLU that determines the intent paired with the LLM that generates the response paired with an NLU that checks the response like you could start to see some interesting frameworks like that to try to get rid of hallucination but I don't yeah it'll be impossible to get rid of the hallucinations 100% of the time and just anytime you're using statistics which is basically what these LLMs are like there's you can never have 100% certainty but it will get very very very very good and you might be able to get rid of certainty for like anything that's too high risk right like if you know if it's like I don't know like a medical bot or something like that anytime you determine it's a topic that's high risk you might actually just choose to use purely deterministic flows if you cannot afford a 0.0001 chance yeah it's a case of how kind of critical the use case is isn't it you know like the risks of the business of getting it wrong and stuff like that if it's involving finances or health or something like that and yeah that kind of rooting use case we definitely put that under the interpreting you're interpreting what someone's wanting then you put them to the right place which is great that makes perfect sense what are your thoughts on the tooling around large language models you're building tooling that that obviously incorporates large language models there's a lot of kind of things around there that harness them one of the areas that is a bit of a gap at least from where I'm sort of sitting is let's take let's take the example there of someone's using a large language model either in that root in case maybe it's not the root one but maybe it's more sort of like more transactional fulfillment based journeys and but it's a company like we did a webinar for example with signer signer express scripts they handle four and a half thousand concurrent conversations so that's like right at this moment in time right now as we're speaking they might be managing four and a half thousand conversations and so you mentioned there the statistical kind of spread I suppose of potential for hallucinations yeah if I was a company like that and I wanted to deploy these models even if it's an FAQ use case if it's an FAQ that's related to someone's prescription and what can I use this drug for versus that drug or whatever you don't really want to get that sort of stuff wrong but there isn't necessarily unless you're kind of either aware or building otherwise kind of like quality assurance tools that can sort of like sit before something goes out to do that kind of monitoring because all I can see is I'm hearing people talk about things like you know what is it the it's chain of thought prompting with what is it called I forget that I forget what it's called but it's basically using large language models to vet the response of various large language models to see where the consistency is taking the most consistent answer and then using that basically and so but that seems to me as though you're using you're marking your own you're grading your own homework you know you're using large language models to vet the response of the large language model so that the limitations of the one technology are inherent in the in the other side I don't know what your thoughts are in terms of that sort of like quality assurance that scale question for for this kind of technology yeah I think sometimes as an industry we again it kind of goes back to like focusing too much on the tooling in the end not thinking about like what's the right tool for the job and just because you know large language models are kind of the thing right now you know like I had a chat with a bank exec and it was it was a pretty funny conversation because it was like we want to use chat GBT and I'm like got it and it was like but I don't want it to hallucinate because it's like this sensitive stuff and I go got it and so it's like you almost want to hand write the messages for chat GBT and it's like yes right and then you sit there and you know you kind of laugh a little bit because you know are we using large language models in places where maybe that's actually not the right place like maybe for sensitive answers you're just going to use an NLU with a deterministic response and that is the correct you know approach to achieve that user experience that you're looking for where you know if the correctness of the answer is the most important thing then maybe that is that is the right approach like I think what's danger about you know you know we've we've all had friends in the past where you know maybe they say something and you're like you don't know what you're talking about it's far more dangerous dangerous if that friend sounds really confident and sounds really right like at least with the bots before they would break and say I sorry I don't know that now they just they kind of you know give you an answer that sounds pretty good I almost argue that that's more dangerous in a lot of scenarios and so yeah I think it totally depends on the use case like we haven't seen this as much like you know it's tough to do a lot of this stuff in the enterprise you know we've you know I think as an industry we've sort of been pontificating on it for like six months but I think what you probably end up seeing is something like an NLU to do the inference just given that LLMs are really expensive with an LLM to generate the response and if it's a net new response you use the LLM to generate it and over time I think you actually start to cache answers you sort of say hey this was an answer that's 100% correct it was generated by an LLM and so over time like you start to see more and more of the actual responses given being sort of if it's in that short head if it's in the normal distribution where it's like the 80% of the most you know the majority of the volume is like 20% of the questions those 20% of the questions are like totally hard coded like you have like variants and stuff but like you know it's a pre-written response and it's only for the things where you've never seen it before you know it's a sort of a net new response you let the LLM take a crack at it and I think you then try to apply QA yeah these things are expensive right now LLMs are not cheap to run in production at like crazy scale so I think you probably only want to use it as like a last you know last resort but otherwise you're going to use like an NLU paired with a deterministic response so a lot of this is under customer support there are use cases for agents were like you know you you've you can't do it like I don't know it's a good example maybe like if you have like tons and tons and like actually drive-throughs are a great example of this drive-throughs require so much variation in your ability to handle conversation the LLMs kind of enable the whole thing to actually be possible like I tried to build a drive-through assistant with NLU and deterministic answers two years ago and it was like it was crazy difficult because they're like you know there's a whole myriad of reasons that's something that's now totally unlocked with LLMs but maybe like traditional customer support which is like FAQs refunds like you know contact you know contacts all that kind of stuff that might actually still be better with traditional NLU and deterministic responses so choose the right tool for the task is my TLDR yeah I think that's that seems fair as well because like the it seems to me at least as though the hype and the moment has kind of got everybody sort of swept up a little bit maybe it's just dying down a little bit now but it was kind of like LLMs are the answers to everyone's prayers whereas really and pragmatically you have to find out where LLMs are actually going to add value rather than create more work and or create more risk you know and so that I think that that's where we'll see them used I think more broadly in enterprises is in places that enhances an NLU for certain processes that require deterministic kind of models and they don't necessarily have to be driving the conversation in order for them to be really useful and effective you know yeah I feel greatly we're on sort of the come down of the hype cycle like you know we at Voicela we were certainly caught right up in it I remember saying to one of our board members you know I believe the AGI was right around the corner and looking back at it now it's like well how I defined AGI you know was was fairly vague but we were like there was a couple months in I don't know call this like like late Q1 where it was like every week it was like GPT-4 came out it can now do images baby GPT auto GPT age it was like it was like if we just you know I was just like extrapolating the curve at a couple more months and I was like holy you know this is going to change change quite a bit and I think now you know the way I view LLMs is a much more tame approach which is these are you know conversational AI has always been the intermingling of the interface and the end user interface which is like you know conversational that kind of stuff with the channel and then business logic and now I think the channel for the most part has stayed the same people are still using like web chat interfaces IVRs like that that hasn't been revolutionized like the actual end channel the interface that's used in that channel has become extremely fluid and it can glide really easily and it can like handle conversations and that's awesome and then the business logic for the most part has not changed and you know the system integrations like integrating with your CRM and you know integrating with your you know agent light you know handoff platform all that kind of stuff that hasn't changed and so when I think about like the three sort of core things that could have dramatically changed it's really right now just that sort of interface in the middle where you can have really fluid conversations that's amazing but ultimately you still have to design a business logic and you still have to have it deployed to a web chat or an IVR or an SMS or something like that.
"braden" Discussed on VUX World
"braden" Discussed on VUX World
"We were able to get to a pretty large chunk of the Fortune 500 now in terms of just sort of customers. We've been super blessed to have the incredible customers that we do. And now with this latest funding round we actually had the majority of our previous funding round still in the bank. And so we're sitting on a lot of cash right now and we're trying to think about how do we actually now do what our customers have been asking for which is two years, which is, hey, you know everyone's been pointing it out. They say Voicela has this incredible conversation design tool you guys have built over like the past like 40 years basically. You guys also have a really strong NLU. You also have an NLU manager analytics and transcripts. Like what am I not getting, right? You sort of have all the end-to-end components. And now we feel like we're in a position where we can start to say, hey, you know what? As you do evaluate vendors, you can use Voicela for just conversation design, use this in our traditional business but now also you can take us end-to-end. So we're having some of those conversations with customers and that's been a really exciting sort of transition where we feel like we're able to. And also, you know, I think a lot for startups if you're in a really big market like Conversational AI you got to be amazing at like, you know at least one thing where you're known for because I think you have a lot of like jackable trades master of none vendors. And that's a really tough spot to be in when you're in such a competitive space. Like you go to these conferences like Voicela and AI in September, there's, you know, my plug for Pete and there's a ton of vendors where it's like kind of all the same stuff, right? Same catch phrases, same, you know metric that they're scoring on. And so we want to be, you know, world-class at one thing and then slowly over time be world-class sort of the whole stack but make sure that we don't lose that initial edge. So yeah, people have been predicting it for a long time, including yourself. And, you know, we're now sort of coming out of the shell a little bit and with this latest funding ground, try to, you know really signal strength that, hey, you know we're going to take this pretty seriously and become the world's best conversational AI platform with a focus on collaboration, being tech agnostic and then having that design centricity that sort of, you know collaboration between designers and developers. Those are really the three things that we're focusing the platform on. Hmm. Yeah. And it makes sense. I mean, I remember the discussions that we've had in the past, which is that whether whether other platforms are kind of jack of all trades or what have you isn't necessarily for me to say but I think that definitely the difference that I can see is that most platforms are very much technical platforms. Like, even though like not like, you know generally speaking, I can open up dialogue flow CX. I know my way around it. I can build something, same with IBM Watson same with dialogue flow ES really. You can kind of get away with it. But the problem is that really to build a real robust conversation in there and I used to see some of the stuff that Justin's been building in dialogue flow CX, you know obviously to adjust and it's like, it's it becomes very complicated very quickly when you have a technical brain that that builds out production grade software. And so those tools always for me of at least have seemed to be technical first tooling whereas voice flow has always been design led. So I suppose, and I don't know whether or not I'm asking you to give away any of the secrets here but like if you look at the platforms like that like dialogue flow and IBM Watson, like Brett presented some really great information that we commissioned him for ahead of Umpires which was when he asked like what design tools or what frameworks do people have experience with and it was like neck and neck between dialogue flow and voice flow and all other vendors were sort of down there which almost to me kind of speaks to the split of skill sets in the industry. You've got loads of designers, loads of developers but there doesn't seem to be that kind of middle ground and this isn't necessarily a criticism of voice flow but one of the things that was that was always a bit of a challenge was that how do you get all of that robust complex design from voice flow into Watson or something like that you know, I know there's export tools and so on but it's like there never seems to be this kind of middle ground between development and design whereas that seems to me to be where voice flow potentially can play so it's the goal from having the easy to use design friendly agnostic kind of tooling and then moving more into the technical sort of the technical space. I wonder whether like I had two questions about that. One is like who do you see your customer being now or becoming and two in terms of the sort of like platform itself where do you see that voice flow needs to improve if it is going to become as enterprise ready as a dialogue flow RBM Watson. So maybe we'll do the second question first and then we'll go to the target audience one after that. Yeah, it's a great question. So you know, I think the way to think about this is you have a developer so developers you know, if they generally know how to code right you know, I don't want to be brought to give too broad sweeping of a statement but generally speaking developers can code. And so when we think about like the tools that a developer wants to use it's often not using a GUI and so you know, the GUI is there to make some functions of like the development process easier, but the actual development is going to be done with code. And so I think what you've had is you've had the developer platforms like a dialogue flow basically say oh we'll do the low code thing and they start from the developer and then they start to add on you know, sort of their their GUI interface but it's essentially a GUI for a development tool. The way that we're thinking about this is let's actually segment the developer platform which is all of our API documentation right. Have a service oriented approach where we are exposing as much of VoiceFlow's platform as we can through great SDKs, APIs, really good developer documentation so you can build custom tools and workflow on top of VoiceFlow and you never have to go into the GUI. And then for the designers let's have an incredible GUI. And so I think it's I think the approach of dialogs on Watson a lot of these sort of like large developer platforms has been to couple them which I think and then you know with as when you're a developer you can kind of do anything that's like the most amazing part about being a developer. The challenge is a GUI cannot have everything in it. GUIs are opinionated and being a developer is actually allows you to sort of have this unlimited amount of workflow. So those two things if you try to combine them you end up having a GUI that has a million buttons and a million switches and toggles and this and that because you're trying to put all the power of code into something that is ultimately a constrained interface. And so the approach that we're taking at VoiceFlow is the GUI which like essentially the design tool remains extremely opinionated. In fact we're making a lot of like big changes behind the scenes right now to make the GUI that people love and use even better. Like how do we continue to double down and reinvest in that system because we you know people are like we love VoiceFlow and we sit there and we see all the flaws we're like oh this could be better this could be you know smoother da da da da da that's going to you know we're going to continue to invest really heavily there. But for the developer platform we don't think we have to add complexity to the GUI we think what we can do is start to showcase all the underlying APIs and we're starting to see this. So last week for example we just launched a brand new API which is the Knowledge Base Management API. What this allows you to do is to upload replace manage all of your knowledge base programmatically through an API. Why that's really valuable is the developer doesn't even have to go into VoiceFlow. You can actually even create your own GUI layer your own internal tooling and you're managing your VoiceFlow system programmatically. And so by bifurcating the platform into these two different areas you're allowing the developers to do what they do best which is use the tools we've given them but then build their own tooling their own layers own workflows around it and then give the designers something that is meant for them. And so we think that's the correct approach and ultimately we think that that's what's going to allow for that simplicity that people love from VoiceFlow but also paired with the scalability since I think you know trying to do it all you end up kind of being like nothing for no one. I think that's where a lot of the platforms like Dialogflow CX have kind of got themselves stuck and then they've tried to make a developer tool into a design tool and you end up you know being a developer platform which you know requires you to use the GUI in some places and isn't very flexible and you have designers being like what the heck is this there's just way too many developer you know toggle switches you know widgets and stuff all over the place. So to your second question that's it. On the first question in terms of how where do we focus now who's our customer? VoiceFlow for the past two and a half years has been conversation design teams that's our customer. Like the actual end user is the conversation designer and then ultimately like the customer is the team the team itself right so we're thinking about workflow and all that kind of stuff. We're still selling the same folks the differences we're going one layer up and that it's not just conversation designers and their teams because that's one part of the broader conversation AI team now we're thinking about product owners you know the actual developers the data scientists how do we give them all a place in VoiceFlow where they actually have a seat at the table you know it's a one collaborative interface and again when we think of collaboration like every single role has its own you know unique way to work because we say we're going to target developers doesn't mean we're expecting to go to the GUI how do we give them the tools to interact with VoiceFlow without even having to go to the GUI how do we give the designer their tools how do we give the product owner a management dashboard so they don't have to you know like how do we give each owner of each different part of the workflow their own place in VoiceFlow and ultimately we think that will become sort of like the ultimate collaboration platform that can span from design to development and all that good stuff so bit of a rant bit of tangent but hopefully that helps sort of give everyone the picture of how we're thinking about the platform which is ultimately build the interface that works best for the end user if it's a developer we're going to give them great APIs if it's the designer we're going to give you the world's best conversation design interface both of them work together seamlessly but ultimately you have your own place you mentioned though that that you have seen the usage double over it might have been double or triple since the launch of ChatGPT so you've got a lot of customers who are like enterprise conversation design teams but I imagine correct me if I'm wrong but I imagine that a lot of the increase in usership over the last kind of 8 to 10 months has been from people who are keen to explore large language models they probably have either experimented with LangChain or AutoGPT or these kind of tools and all of a sudden now there's a system that does kind of a similar thing but requires a bit less coding but the community I suppose that's building up around large language models is more of a technical community rightly or wrongly it's the people building the open source tools they're the ones kind of like building all of this kind of ancillary MLOps kind of tooling would you say that that's the people who've been joining Voiceflow recently like not from the enterprise side but more from like the hobbyists who want to just get in there and test it out as a tool like would you think would you because I would assume that you've probably had more development or technical people using the platform over the last over the last few months way more right and I think you know you've guessed correctly that a lot of the growth that we've seen has been that sort of technical audience and that's been pushing us further in sort of this direction of hey how do we maintain the best conversation design experience but also service these other folks who want to customize they want to tweak they want to use you know custom interfaces and things and we've really been trying to be the bridge between these two communities and ultimately we think what's going to be the most powerful thing is how can Voiceflow enable these incredible developers who are building you know I saw we have a big discord community that's blown up over the past like couple of weeks and I saw this one developer who created two things they created a custom web chat widget where they're like I want it to like you know be waving to the customer in the bottom right and I wanted to you know have all this like cool you know like Calendly functionality where like you know you basically book book meetings through the the widget and they just built that themselves using our open source and then now they're able to share that with designers and I thought that was really cool like how do we enable the developers to actually build and give value so you know that's what being a true platform is and we're trying to sort of push in that direction how do we allow customers to actually provide value for each other as well right so you know we've seen this over the past couple years people who know Voiceflow you know it's become a bit of an employable skill you see in job descriptions and stuff that's great that's you know value between the designers now how do we allow the developers to give value to the designers and vice versa right how do we allow them to have a single platform where they can work together create things and ultimately share and so I think that's been really exciting and we're going to see how that continues to sort of emerge and you know we've got some fun features coming up which we think are going to even supercharge this further for the community but ultimately yeah it's that mixing of design development which was where we actually think like sort of the sweet spot and like where the magic happens is because Voiceflow has largely been designed for the past like two and a half years so to see this influx of developers who are able to extend Voiceflow even further is awesome here's here's a by the way this is my favorite example we had a developer who they saw our knowledge base and they go you know it's actually we you know we don't like the interface that Voiceflow provides and so they took our API and they built their entirely own custom knowledge based interface on top of Voiceflow that they've now been giving to their their clients which we thought was awesome right and you know it's funny at first like I was like hold up you know I'm offended you don't like our our knowledge based interface but then the best part about being a platform is you know if you don't like how Voiceflow presents the information or if you want to tweak it a little bit you want to add a block that we're missing you want to change the prototyping experience change how the chat interface looks like we should give you the ability to do that because you know we will never be perfect for every team but if you have the ability to develop we can now be perfect for every team right and hopefully you know customers start to help work you know work with each other and you know fix those gaps and things and so we're super excited to see that sort of like emerging the emerging platform ecosystem which is going to be really cool definitely I remember I remember listening to a podcast a long time ago with Ben Thompson from Stratechery and he defined platform as a suite of capabilities upon which further value can be added so like Windows as an operating system is a platform because on top of that you can build an application yourself that then adds further value to the whole platform now we've got Word so now everyone can use Word as further value to the whole thing so that's kind of and when a lot of the time when people say that they have a platform most of the time it is kind of an out of the box suite of tools and yes you can assemble them to add value but when it comes to extending it for a very particular case that you need to extend it for like that knowledge base example that you give that's when you find actually a lot of the platforms are closed and you can only do what you can do within the confines of the environment you're given so I definitely like that approach I'm curious about we've got a question here from Oya which I'll come to in a minute because we kind of touched on sort of like multi-modal that widget you mentioned the Calendly widget and stuff like that but I'm curious about sort of like thinking about the future there's two maybe there's two paths I can see but maybe you can go down both paths one is continuation of providing tooling for the enterprise which is for conversation design teams and development teams trying to bridge that gap trying to kind of like you know have them enhance use large language models responsibly and stuff like that but then at the same time you've got a lot of hobbyists and these are the people who are building your likes of the Langchains and they're building applications with that and so potentially the other side of this which whether it's more scalable or not you tell me is that Voiceflow being used potentially as the underlying platform that sits underneath future businesses that are kind of like AI first businesses and so I don't know if you've given any consideration to because it seems as though those influx of new users who are the developers they're not necessarily working at an enterprise trying to build applications for those they're probably trying to figure out okay how do I build something that's going to let me build a business around or something like that would you say that's kind of among the right lines? No totally and like so I'll touch on like the general sort of framework we're thinking about this and then I'll go specifically into multi-modal on the general framework I think there's sort of three types of like software companies essentially there's tools and this is like sort of in the B2B space there's tools which is there's Voiceflow and you know we build a tool which provides value to solve a particular business problem you know and we give you value and we hope that you pay us money and you know that's sort of that that's the business and then there's a platform and a platform is which we give you as you mentioned a set of capabilities of which you can extend further so a tool is going to be just sort of you know you have an accounting problem or accounting software now we're saying hey here's a bunch of building blocks of which to build to build even further value upon right so that's like sort of what a platform is I think the final step is when there's an ecosystem and that's when it is you know the platform and then you have users who are able to without knowledge of the platform or without even needing the platform they're providing value to each other and so a good example of that is going to be like let's say Shopify consultants for building out you know different Shopify sites different Shopify templates you know you can see Salesforce there's a whole ecosystem built around you know the implementation and consulting for Salesforce and I think you know as we think about AI agents the reason that we haven't closed Voiceflow down like you know we have a ton of users who are just on a free plan is because we really want to build that ecosystem how do we allow for users to build on top of Voiceflow build businesses on top of Voiceflow where folks are building consultancies they're able to build implementation services training services I've told folks before like you know the folks at CDI who are awesome you know we don't want to be the company that's you know doing everything for every customer we'd much rather have value circulating around right we just want to focus on how do we build the best possible platform and then how do we enable the community to provide value to each other and you know sort of build on top so that's where we're going to just continue to extend or open more and more APIs so that Voiceflow you know let's say there's Voiceflow has a conversation design platform a knowledge base an NLU and it has a dialogue manager right let's we have more than that but there's just four you know platform components let's say customer likes A, B and D but they don't like C right let's say they really prefer dialogue for NLU we want that customer to be able to say okay you know I can still use Voiceflow I'm going to build with A, B, C and D but C is going to be a different vendor like how do we allow people to swap to build out their ideal workflow and technology stack we think that's what's like sort of the most exciting opportunity is to be able to sort of build that like ecosystem where people are helping each other do that and a platform that allows for that to be possible and then last on multimodal so what you're going to see is we're trying to lead with how do we allow everything to be possible in Voiceflow first and so this has been our approach for years now how do we allow for almost like a hacky solution to get done so when you see like me posting LinkedIn videos and things it's often what is like the MVP approach where it's possible now you can do it it might not be well productized though it might not have a great GUI experience it might not have a good UX and like over time as things become more and more core to Voiceflow and like to all users we'll make that increasingly core so a good example of multimodal is we actually have we have a couple customers who are using a POC that we built which was touch hotspots so the ability to have like a phone with a prototype that's actually going to go ahead and like you can use touch you can use voice it's like this fully sort of multimodal touch prototype and if enough customers get value from that then we start to think about well how do we really productize this how do we make it a really seamless UX inside Voiceflow instead of something that you're just building on top of the platform and so I think all yeah to your question on multimodal a lot of this stuff is possible in Voiceflow today but it requires some of the developer sides of the platform and over time as we see more demand and people are starting to like use some of these platform features we'll look to actually sort of verticalize inside of Voiceflow and make it a core part of the experience but today we're actually just laser focused specifically on web chat the dialogue API and the actual just platform itself those are the those are the areas we're really really focused on because if we try to do everything all at once we're going to kind of give you guys like the luke water you know like lukewarm platform experience where it kind of you know it's a jack of all trades master not master of none we want to give you guys like the best collaborative conversational AI platform it's by far the best and then over time expose more and more API's so that you guys can build what you need first and then over time you know hopefully Voiceflow becomes a bigger company we're going to be able to go ahead and extend so yeah hopefully that was a helpful framework to see how we make decisions yeah absolutely absolutely we've got a question here apologies because I can't see it just comes through as LinkedIn user for some reason maybe it's because we're not connected or whatever but it's it's someone who's saying are we interested in platforms or frameworks I think we get power from frameworks so what in your kind of opinion is the difference between a platform and a framework great this is great so in my opinion the platform is the ability to do something and the framework is the best way to do it that's layered on top and so what we're trying to do is build a technology platform that makes kind of anything possible and then we're layering on our opinions so what's a good example of that in Voiceflow you have domains topics so essentially like what's the information architecture of the agent itself right how is it structured the platform doesn't really care per se like it's all you know it's you can build you know transformers to turn Voiceflow's JSON into another into another structure you can turn Voiceflow's JSON structure into a dialog flow structure like you can sort of convert it all over the place but we started with an opinion and that opinion is sort of the Voiceflow framework of which like all of our standards are built on top of it so for us it's called the .vf file which is the .voiceflow that then leads into the GUI itself is all structured and that's that same sort of thinking which is domains are like the collection of sort of it's the broadest possible grouping of functionality it's like you know everything to do with the refunds and then you go into topics where it's like what are the specific types of things that we're doing within refunds and then you go into flows which are like the very specific like multi-step interactions in order to complete that that's our general Voiceflow framework but it all sits on top of a technology platform that is much less opinionated and more malleable so if you want to change that at least you know what our opinion is first and that's what allows you to convert our opinion into another opinion because you need to know how Voiceflow thinks about data it can't just be this like this you know amorphous blob we so yeah hopefully that's helpful we have a technology platform and our framework of how we think about the world is layered on top but because it's a platform it's flexible for you to ultimately work with it and we you know we document our opinion nice nice perfect so out on par what was really interesting is that we had a lot of great content from a lot of seasoned conversation designers who are were all presenting every second talk I would say had some feature of large language model or generative AI in there something like that which is really great to see so you've got you've got the conversation design community seemingly if you were to go on the face of of the content out on past really embracing large language models at the same time though there seems to be and I don't know if this died down now or not but there is definitely this kind of I suppose period of uncertainty where some conversation designers are thinking like well what's my job going to be now you know and then you have kind of like and I've always actually I've always believed that language understanding is first and foremost a technical problem but creating experiences is first and foremost a design problem so you kind of got you have to have both of them working together and when if everything's very much heavily design focused and there's no technical ability you're going to end up with something with no substance that can't deliver on what it promises if something's really technical first with no design thinking it ends up being really clunky and experience is terrible a really good example of that is I won't call out the specific train line but there was Alexa skill and it was a train booking skill and as we were starting to learn conversation design during the course of building Alexa skills you mentioned earlier on situational design situational design being the ability to design for a given state of a conversation and having the flexibility to when someone wants to change the state of that conversation being able to go with the flow and do that so if you're in the middle of booking and checking train times for a train and all of a sudden you want to do something else like ask a question about whether you've got disabled facilities or whatever then you can do that without affecting the state of the flow that you're in whereas and but those experiences that were done which was more of a technically led experience in this train booking example you chose at the beginning you can either check train times or you can book a train if you go down checking train times now all of a sudden you're in this path you know you're down this kind of like tree for whatever the better phrase and you get to the point where you found out your train time if you said okay I want to book that it would never let you do it because you weren't in the right tree you weren't in the right branch for that you had to go back to the beginning go and book and then go through select your times again then you get to you so that was for me that was an example of a solution that was built which was technical first how can we do this well we need train times we need bookings but the experience was terrible and so it feels to me as though we're going through this period with generative AI where really to use generative AI aside from what Voiceflow have built which and some other tools which have started to democratize access to that predominantly these are just API's so you need to have technical prowess to be able to use them in the first place and so people are building tools like Langchain and AutoGPT all this kind of stuff and they're all very technically minded folks and so I think what some designers have felt like is that well large language models have got tremendous power because they can produce dialogue which is part of my job they can understand dialogue which is another part of my job and it seems as though the only people who are building this stuff is all technical people so is this now becoming more of a technical job rather than a design job and so I wanted to ask this on behalf of those conversation designers that might be feeling this, my opinions, these are not my opinions but this is just, I think it's good to get the question out there which is for the conversation designers thinking that actually large language models are a very technical thing, it doesn't really concern me too much, what am I going to be doing and that is my kind of role in jeopardy so to speak, I wonder what your thoughts are on that and how large language models are affecting or impacting the role of a conversation designer or conversation developer.
"braden" Discussed on VUX World
"And so congratulations on your latest round of funding. That's great news, very much a vote of confidence in terms of what you're doing there. You've done a lot, not just since we last spoke, but in the last eight months, to be honest, in terms of iterating the product, every single second day there's a video on LinkedIn with a new feature and stuff like that, so you've really sort of grabbed the ball by its horns. Describe to us a little bit about what that journey has been like and how you're thinking about large language models as part of the, not necessarily the voice for a product, but as part of the tool set for developing and creating these experiences. Yeah, so when ChatTBT first launched, like a lot of folks on this call, we were A, blown away, and B, didn't know exactly how this was gonna play into our business. And the approach we decided to take was, let's innovate really, really quickly, and basically let's learn with our customers. And so what we did is we actually segmented Voicelo's product into two products. So what we used to have was just Voicelo, which is a conversation design platform, and it would sit on top of traditional NLU providers, so like Dialogflow, IBM Watson, Raza, all these kinds of things. And what we did is we partitioned the product and we said, this is gonna be the AI playground, essentially like the AI builder, and this is gonna be our traditional NLU product. And what that allowed us to do was to really quickly start to innovate and just throw stuff at the wall on the AI builder. And we had a legal disclaimer, and it's still there, when you start using Voicelo's AI builder, hey, just wanna let you know, large language models, if you're gonna push it to production, know that they can hallucinate. If you're gonna be playing with different models, there's experimental features in here, all this kind of stuff. And we basically gave ourselves the liberty to try to iterate as fast as possible because we didn't know what was gonna work. So we basically mapped out all the possible use cases that we could initially think of. And what's crazy is we would not even be thinking about transcript management and different examples of prompt chaining and responding from memory, saving LLM responses into variables, all these kinds of things that we've launched in the past couple of months. We wouldn't have been able to predict them six or eight months ago, but what happened is if we were launching features, customers would come and say, hey, what about this? Could I do this now? Could I do this? And only as we launched more features did we understand more about large language models and how they play into conversational AI, which then led the roadmap for the next set of features. And so we've just been trying to take this iterative approach of let's innovate with our customers. Let's try to launch weekly. And that's been really successful. We've seen a ton of usage and people have really been loving the new features. We're actually very close to the point of where generative AI features are about to surpass the original business. It's like every single week it now teeters. They're like neck and neck. And so that's been really exciting to see. Interesting. And so in terms of the genesis of Voiceflow, you began as essentially a story building, interactive story building platform for Amazon Alexa. And then quickly realized that you have something here that can be used and applied elsewhere, such as broadly speaking conversational AI. And then it became more of a design platform where a lot of teams were using it to collaborate on designs and using it to prototype and test early iterations and stuff like that. And although there was always a feature there which let you publish a chatbot, for example, on a front end, it always struck me as though the value and positioning of the value from Voiceflow's perspective was all around teamwork, design, concepting, prototyping, getting your ideas out there so that when you come to commit to building something in Watson or Dialogflow or wherever, you've worked out the kinks, you've got everything, you've got your ducks in a row and there's no surprises kind of thing. Whereas now there's a lot more features regarding LLMs and a lot of the tooling and stuff that you've harnessed around it, it's not necessarily, it seems to me at least, applicable or the same use cases don't necessarily apply where you would do all of this work to craft this experience using LLMs and then just take that and go and build it somewhere else. And so we've had some conversations about this in the past and I was excited to see in the press release that it's explicitly stated there that Voiceflow now sees itself as being a production-ready enterprise platform rather than a design tool. So I'm wondering if you can talk us through that kind of transition. Yeah, so the reason that we chose collaboration as sort of the focal point is we really wanted to serve enterprise product teams and in specific that relationship between conversation designers and their developer counterparts. We really want to be sort of the genesis between the two of them. And we felt that we weren't ready to take on larger platforms like a Dialogflow or IBM Watson. We were just too small a company. And when we looked at our resources, it was let's be really amazing at one thing. Let's be incredible at conversation design. Let's really nail collaboration. And so we spent like two and a half years basically perfecting the conversation design workflow. How do you design, prototype, collaborate, and essentially thinking about conversational AI not just from the outcome but from actually optimizing the inputs of the workflow. Basically saying, hey, a lot of platforms they focus on like what is the art of the possible if you use every single nook and cranny and what's that final widget look like. But then they don't really think about much about the actual creation experience themselves which makes it then hard to achieve that outcome. If you're super outcome oriented and you don't think at all about the inputs and like how people actually get there. Sometimes it can be, it's like if you have the, I don't know, the ability to have the fastest car in the world but it requires so much gas that can only run for a minute or something like that. Like it's actually like, yes, you can achieve that incredible outcome but the inputs matter a lot. So we focus a lot on inputs for two years. And that led us to get to 130,000 global users.
"braden" Discussed on VUX World
"Love it. It's good to see Paul. I know Paul from way back in the Alexa days and doing situational design, if you remember that. Yeah, yeah. Yeah, Paul and I spent many hours on calls together, so that's good to see you. Definitely, definitely. And around about that time, we were on the show. It must have been towards the back end of the Alexa sort of craze, I suppose, was the last time, maybe as you were on, and it was all around. We were talking about discoverability, if you remember that age-old conversation about getting Alexa skills discovered. It was good times back then, but it feels as though we've got another community thriving now. That community is blended with other communities, and now we're kind of as a community growing. The technology is getting better. I don't know about you, but it feels as though we're in a similar sort of space now with large language models, generative AI, similar kind of space as we were with Alexa. Yeah. You know, it's funny, and every time I come on VUX World, it feels like we're sort of at another inflection moment. You know, Alexa was really sort of the first wave of I think this community that sort of came together, and really what it was, it was the conversational AI community around building agents, and Alexa happened to be the platform that we were all using. And then I think we all transitioned into sort of conversation design more broadly, just conversational AI. We started to pick up folks from the contact center space and other places. And then more recently now, it's all about large language models, and that's been sort of the craze for the past six months. And I think what's really cool is this span and scope of large language models in the applications is tremendous. You know, we were sort of big believers in it early on, and we've seen Voiceless user base essentially double to almost triple from the beginning of this year till now. And so that's been really exciting. Introducing folks to conversation design, introducing folks to sort of the conversational AI space, you know, was just so much interest from ChatGPD pulling people in. So yeah, it's been exciting. Nice. Nice one. We've got Lawrence again from Austin. We've got Khadija from London. Hello there. Thank you for tuning in. Feel free to stick any questions that you might have throughout this conversation in that chat, and if we can do, we will definitely try and get to it. Yeah, you're right there. I mean, there is an inflection point that seems to emerge every time we chat. I think one of them was the last time was around how, I suppose it was, as I said, towards the tail end of the Alexa sort of phase when I think a lot of people were a little bit kind of like, they've been knocking on the door for a long time. There were certain things that still hadn't kind of materialized and stuff like that. And so it was kind of like the last few podcasts we did around Alexa specifically was very much kind of focused on that. And it's really funny because I was looking back. We've been transcribing all of our podcasts and actually using Voiceflow to mock up some examples of what the VUX bot would be like. Once we've accurately transcribed all the podcasts and all that kind of stuff, you're going to be able to just talk to the podcast basically and have a conversation with all the guests that were on it and mine, all the information that's been in there, which is going to be great. But I was going back through the episode that I did with Georgia Quinter, and it was in 20, it must have been 2018. And he's talking about three things that were an issue, which was one is discoverability, the other was retention, and the other was monetization. And it just got me thinking, preparing for this conversation and thinking about this conversation, it's like we had that exact conversation probably two years after I had the conversation with Joe and those challenges were still there. Whereas now, I don't think it seems as though, I mean, obviously we can get into some of this stuff around the perception of hallucinations and potentially also the perception of a slight lack of control perhaps if you're using large language models, but at least it feels positive. It feels as though the challenges that may exist are kind of surmountable as you start to mix and blend different technologies together. So it feels like a really positive time to be in this space, doesn't it? Yeah, I think it was interesting. When we were an Alexa skills-focused company, I had a conversation with an investor actually, who goes, you as a tool maker can only make money if your customers are making money. And that was a really profound thought, like you might have product market fit, you might be the world's best Alexa skill building tool, but unless your customers are making money and they're, let's say that's an agency, so voice those cells, tools to people who are building agents, if those agents then are providing value to their end customers, the value has to flow up, right? If the value doesn't flow up, if Alexa skills can't be monetized well, then the Alexa skill builders aren't making enough money, which then makes the tool builders not be able to have enough money, and then it sort of all just flows up the chain. I think what's really been exciting about large language models is we are seeing a ton of people make money now. Whether they're just automating far more conversations than they could before, they're able to hit higher deflection rates, they're able to provide better CSAT scores, or even what we're starting to see too is the emergence of a really strong agency layer servicing SMBs and mid-markets where normally to stand up like an enterprise chatbot, it's really expensive, right? You need a ton of people, full-time staff, all this kind of stuff, but with large language models, it's brought the floor down of cost creation, or the cost floor down is so low that now these one, two, three person agencies are able to service really large companies with really sophisticated bots. That just wasn't possible before, and so the value's now flowing up, which has been really exciting to see. It's a big change, I think, for this community to now have use cases that are like, there's no more opportunity ahead of us than there is behind us at this point, which I think maybe wasn't the case with previous channels.
"braden" Discussed on VUX World
"All right. Hello, ladies and gentlemen, boys and girls. Welcome to VUX World. I was going to say, all right, all right, all right there. So I've kind of been slipping into and someone pointed out that it's a Matthew McConaughey quote. And so for some reason I just slip in. All right, all right, all right. Hello, welcome. But hello and welcome to VUX World. Today you'll notice I am wearing my trusty Voiceflow cap because today is Voiceflow CEO Braden Ream's third appearance on VUX World. We go way back and today is going to be an epic conversation. We're going to be discussing large language models and the future of conversation design and the future of AI application design and the future of Voiceflow. I'm looking forward to that coming up in just one minute. But before that, I need to give a shout out to our presenting sponsor today, Tideo. Tideo is a conversational AI platform that specializes in catering and providing for small to medium sized retailers. So if you are a retailer, maybe you're running on something like Spotify, maybe you're running on something else, and you want to look at ways in which you can utilize artificial intelligence, then you should check out Tideo. It's got a live chat platform plus conversational AI capabilities. It's got a lot of pre-built intents, pre-built conversations around order check-in, product availability, shipping status, returns, all that kind of stuff. And it is answering somewhere, what is the figure? It's four out of five customer questions it's answering successfully. So if you want to benefit from artificial intelligence, but you don't want to put in a whole load of groundwork, time and effort, and you want to get to market a little bit faster, then you should check out Tideo. You can, if you want to, go to tideo.com forward slash V-U-X and there you will find a promo code where you will save 20% if you were to go and join today. That is Tideo, T-I-D-I-O.com forward slash V-U-X. Thank you, Tideo, for partnering and sponsoring V-U-X World. Now the other thing we definitely need to discuss is, it's coming around so thick and fast, I can't believe how fast it's coming around. It is the Voice and AI Summit in September. It's at the beginning of September from the 7th, sorry, from the 6th through the 8th, and it is going to be absolutely immense. We are going to be in Washington, D.C. for the Voice and AI Summit. For those of you who are familiar with the Voice Summit, you'll be familiar with the event. It's been slightly rebranded, it's now focusing more on generative AI, large language models, that kind of stuff. Last year we had a track there, we had a stage, it was absolutely immense. We were focusing on enterprise application of AI, and today, not today, but in two weeks' time, we'll be doing the exact same thing. The V-U-X stage will be back and we'll be looking at how large enterprises, including some large banks, including some big insurance providers, including some big healthcare companies, how are they utilizing this emerging technology in the enterprise today? And if you were to do something similar, how can you do that? Effectively, responsibly, and with a high degree of quality. Now, speaking, you can go to voiceand.ai, the website is, and you can get your tickets to find out more there. So speaking about utilizing large language models effectively, let's welcome on today's guest, Braden Rehm. Voiceflow have just raised another round of funding, another $15 million to put into growing and scaling the platform, and so I'm delighted for Braden to join me for the third time on V-U-X World. Welcome. Well, welcome myself. Yeah, excited to be back, third time. Wow, Patrick. And now I'm wearing the Voiceflow hat, which I was lovingly given at UNPERS in July. Thank you for partnering with us for that event. It was absolutely fantastic to see you and the team over here in the UK, and to support the first in-person conversation design conference in the world, which was absolutely fantastic. And now I owe you one of these for your hat trick. Your third appearance on V-U-X World gets you the yellow V-U-X cap. Very exclusive merch, this, Braden. So I'll be bringing this across the pond to Washington to present to you at the Voice and AI Summit. So there you go. Love it. Yeah, no, it was an awesome conference. Thank you for putting on UNPERS. We're excited to be back next year. Definitely, definitely. It was really good, actually. I don't want to overly plug our own events and stuff like that, but it did have a very nice communal feel about it. It was really great content, good vibes, good afterparties, which are also very important. And yeah, looking forward to it. Can't wait to have you along next year. We have some people tuning in. We've got Paul Cutzinger. Shout out to Paul. He says, hello, party people. Jose Burgos, how are you from Austin, Texas? Hello, how are you from Austin? There you go. Hi from Portugal. Elson, welcome from Portugal. We are truly global-ridden right now.
A highlight from UX Design with LLMs, with Braden Ream, CEO, Voiceflow
"All right. Hello, ladies and gentlemen, boys and girls. Welcome to VUX World. I was going to say, all right, all right, all right there. So I've kind of been slipping into and someone pointed out that it's a Matthew McConaughey quote. And so for some reason I just slip in. All right, all right, all right. Hello, welcome. But hello and welcome to VUX World. Today you'll notice I am wearing my trusty Voiceflow cap because is today Voiceflow CEO Braden Ream's third appearance on VUX World. We go way back and today is going to be an epic conversation. We're going to be discussing large language models and the future of conversation design and the future of AI application design and the future of Voiceflow. I'm looking forward to that coming up in just one minute. But before that, I need to give a shout out to our presenting sponsor today, Tideo. Tideo is a conversational AI platform that specializes in catering and providing for small to medium sized retailers. So if you are a retailer, maybe you're running on something like Spotify, maybe you're running on something else, and you want to look at ways in which you can utilize artificial intelligence, then you should check out Tideo. It's got a live chat platform plus conversational AI capabilities. It's got a lot of pre -built intents, pre -built conversations around order check -in, product availability, shipping status, returns, all that kind of stuff. And it is answering somewhere, what is the figure? It's four out of five customer questions it's answering successfully. So if you want to benefit from artificial intelligence, but you don't want to put in a whole load of groundwork, time and effort, and you want to get to market a little bit faster, then you should check out Tideo. You can, if you want to, go to tideo .com forward slash V -U -X and there you will find a promo code where you will save 20 % if you were to go and join today. That is Tideo, T -I -D -I -O .com forward slash V -U -X. Thank you, Tideo, for partnering and sponsoring V -U -X World. Now the other thing we definitely need to discuss is, it's coming around so thick and fast, I can't believe how fast it's coming around. It is the Voice and AI Summit in September. It's at the beginning of September from the 7th, sorry, from the 6th through the 8th, and it is going to be absolutely immense. We are going to be in Washington, D .C. for the Voice and AI Summit. For those of you who are familiar with the Voice Summit, you'll be familiar with the event. It's been slightly rebranded, it's now focusing more on generative AI, large language models, that kind of stuff. Last year we had a track there, we had a stage, it was absolutely immense. We were focusing on enterprise application of AI, and today, not today, but in two weeks' time, we'll be doing the exact same thing. The V -U -X stage will be back and we'll be looking at how large enterprises, including some large banks, including some big insurance providers, including some big healthcare companies, how are they utilizing this emerging technology in the enterprise today? And if you were to do something similar, how can you do that? Effectively, responsibly, and with a high degree of quality. Now, speaking, you can go to voiceand .ai, the website is, and you can get your tickets to find out more there. So speaking about utilizing large language models effectively, let's welcome on today's guest, Braden Rehm. Voiceflow have just raised another round of funding, another $15 million to put into growing and scaling the platform, and so I'm delighted for Braden to join me for the third time on V -U -X World. Welcome. Well, welcome myself. Yeah, excited to be back, third time. Wow, Patrick. And now I'm wearing the Voiceflow hat, which I was lovingly given at UNPERS in July. Thank you for partnering with us for that event. It was absolutely fantastic to see you and the team over here in the UK, and to support the first in -person conversation design conference in the world, which was absolutely fantastic. And now I owe you one of these for your hat trick. Your third appearance on V -U -X World gets you the yellow V -U -X cap. Very exclusive merch, this, Braden. So I'll be bringing this across the pond to Washington to present to you at the Voice and AI Summit. So there you go. Love it. Yeah, no, it was an awesome conference. Thank you for putting on UNPERS. We're excited to be back next year. Definitely, definitely. It was really good, actually. I don't want to overly plug our own events and stuff like that, but it did have a very nice communal feel about it. It was really great content, good vibes, good afterparties, which are also very important. And yeah, looking forward to it. Can't wait to have you along next year. We have some people tuning in. We've got Paul Cutzinger. Shout out to Paul. He says, hello, party people. Jose Burgos, how are you from Austin, Texas? Hello, how are you from Austin? There you go. Hi from Portugal. Elson, welcome from Portugal. We are truly global -ridden right now.
Kreider helps Rangers beat Devils 5-2 to force Game 7
"The range is forced to decisive game 7 by beating the devil's 5 two in game 6 of their opening round series. The Devils led one nothing late in the first period, but the range is tied in the final minute when Chris quite a redirected meek is the bandages shot on the power play. The bandage ad then scored his first of the series midway through the second period to give New York a lead it never relinquished. The winds are all the matters and we got the win that we needed and wanted tonight and now we just keep going. Vladimir tarasenko Barkley gadol and Braden Schneider also scored for the rangers who got 34 saves from E grocery and gold. Game 7 Monday in New Jersey. Time area of New York.
Point scores 2, Lightning snap skid with shutout of 'Canes
"Andre vasilevskiy made 31 saves in net and Braden point scored two goals as the Tampa Bay lightning shut out the Carolina hurricanes for nothing. Hey coach John Cooper said the team earned a win after a tough weekend. We didn't get any points out of the Boston game, but I thought we took some steps forward, and I had the guys got rewarded tonight. School is heading into the second period, the lightning got goals from Steven stamkos and point to take a two zero lead through 40 minutes. Point would add another in the third period before Alex killorn's empty net goal would seal the win. I'm Dennis Cox.
Capitals' playoff hopes take another hit with loss to Blues
"The capitals playoff hopes took a hit with a 5 two loss to the blues. Sammy blay scored twice and Joel hofer made 33 saves in his season debut. I was obviously a lot of uncertainties with being in being the a all year and first game in a long time, but I thought the guys put really well in front of me and kind of kept in the outside and like I said earlier, super nice to get the one. Kasperi kapanen and Jordan Cairo also tallied and Braden Shen edited an empty netter. Martin Ferrer vari and Nicholas Baxter had the third period goals for the capitals who are 5 points out of an Eastern Conference playoff birth with 12 games remaining. I'm Dave ferry.
"braden" Discussed on Hay House Meditations
"What do you <Speech_Male> <Advertisement> want me to <Speech_Male> know? <SpeakerChange> <Speech_Music_Male> Right now. <Music> <Advertisement> <Music> <Advertisement> <Speech_Music_Male> <Advertisement> And <Speech_Music_Male> <Advertisement> your heart will answer you <Speech_Male> <Advertisement> very, very quickly. <Speech_Male> <Speech_Music_Male> <Speech_Male> If there's a long rambling <Speech_Male> answer, it may <Speech_Male> <Advertisement> not be your heart that's <Speech_Male> <Advertisement> responding. Go <Speech_Male> <Advertisement> back, touch <Speech_Music_Male> <Advertisement> your heart, <Speech_Music_Male> <Advertisement> <Speech_Music_Male> <Advertisement> restate the question, <Speech_Music_Male> <Advertisement> my heart. <Speech_Music_Male> <Advertisement> <Music> Do <Speech_Music_Male> you want me to know <Speech_Music_Male> right now? <Music> <Music> <Music> The <Speech_Male> second question. <Music> <Music> <Speech_Male> Second <Speech_Male> question, my <Speech_Male> <Advertisement> heart. <Speech_Music_Male> <Advertisement> <Speech_Music_Male> <Advertisement> What do you <Speech_Music_Male> <Advertisement> need <Speech_Music_Male> <Advertisement> from me? <Speech_Music_Male> <Advertisement> Right <Speech_Music_Male> <Advertisement> now. <Music> <Advertisement> <Speech_Music_Male> <Advertisement> My heart, what do <Speech_Male> <Advertisement> you need from me <Speech_Male> <Advertisement> right now? <SpeakerChange> <Music> <Music> <Music> <SpeakerChange> <Speech_Male> And that question <Speech_Male> <Speech_Male> often slows <Speech_Male> down the answer, your <Speech_Music_Male> <Advertisement> answer is <SpeakerChange> right there. <Speech_Music_Male> <Speech_Music_Male> <Advertisement> <Speech_Male> <Advertisement> Typically comes in a <Speech_Male> <Advertisement> single word, <Speech_Music_Male> <Advertisement> a brief phrase, <Speech_Music_Male> possibly a <Speech_Male> sentence. <Speech_Male> Because this is <Speech_Male> the way the heart responds. <Speech_Male> <Speech_Male> <Advertisement> The <Speech_Male> <Advertisement> heart does not <Speech_Male> respond. <Speech_Male> With a lengthy, <Speech_Male> rambling, <Speech_Male> <Speech_Male> dialog, trying to <Speech_Male> justify its answer, <Speech_Male> that is an <Speech_Music_Male> ego answer. <Speech_Music_Male> <Speech_Male> <Advertisement> And if you're receiving <Speech_Male> that, simply <Speech_Music_Male> <Advertisement> go back, <Speech_Music_Male> <Advertisement> reconnect with <Speech_Music_Male> <Advertisement> your heart, touch your <Speech_Music_Male> <Advertisement> heart and restate <Speech_Music_Male> the <SpeakerChange> question. <Speech_Music_Male> <Music> <Advertisement> <Speech_Music_Male> <Advertisement> <SpeakerChange> <Speech_Male> If you're listening to <Speech_Male> <Advertisement> this broadcast, what I <Speech_Male> know is that many <Speech_Male> of you <Speech_Male> are making big decisions <Speech_Male> in your lives right <Speech_Male> now, jobs, <Speech_Male> relationships, <Speech_Male> careers, <Speech_Male> <Speech_Male> health, <Speech_Male> <Advertisement> <Speech_Male> and their <Speech_Male> decisions, <Speech_Male> everyone you <Speech_Male> ask gives you a different answer <Speech_Male> and there's no one to <Speech_Male> turn to. You <Speech_Male> can't Google the answer. <Speech_Male> <Speech_Male> You know, everybody <Speech_Male> <Advertisement> has a different opinion, <Speech_Male> and that can <Speech_Male> be very confusing. <Speech_Male> <Speech_Music_Male> It's been invited <Speech_Male> to take the space <Speech_Male> you've created right now.
"braden" Discussed on Hay House Meditations
"And release. I didn't hear. It release. Inhale again. And release. And if you can continue breathing, just about that pace. Just allow that rhythmic breath. And when you breathe a little
"braden" Discussed on The Paul Finebaum Show
"Second hour Nashville as Alabama wins game one. 70 to 49, Tennessee playing Missouri right now with a slight lead. A couple of games after that, Braden Gaul joining us from ESPN radio. Braden, it is a good see. I feel like you know this audience pretty well. You've sat in spring and summer and been subjected to all the things that subjected. What are you talking about? It's a pleasure. It's a pleasure to see you. Welcome to Nashville and hosted the final I'm sure I have to go through. These are our people, Paul. They are also involved in a company called four 44. What is that? Local digital company here in Nashville, all natural sports. So if you care about things like the predators or the Titans, SEC football, of course, we got you covered four 40 sports on the Twitter account there on the YouTube page, the whole deal. So we believe in giving people a communal experience, which I think is what the fine bomb show is all about is this sort of unifying togetherness and love of a common command deal and that's what we've got at four 40 sports for Nashville. And sort of middle tennesseans as well. But it's all digital. Check it out. We appreciate it. I've always been curious about Nashville because I remember Nashville and it didn't have anything. And the Titans and the predators, it seems like especially the predators have really grabbed this town. That's true. Football still dominates. If you look at the TV ratings, it's still absolutely the Titans and SEC football. In fact, what's fascinating is that the vols and Bama have battled for supremacy in the city. Really? Yeah, so I moved here before the predators in the Titans came here. I moved here in 96 and it was a Tennessee town. Four to Tennessee was the SEC east championship game. It was the SEC championship game. That's the rivalry I grew up with in high school here. A lot of power teas everywhere in Nashville it was a Tennessee market. Right around 2007, Paul, I don't know what happened around then, but right around 2007, there are a lot of more script days, a lot less power tease in the market. A lot of Alabama fans in Tennessee ratings in middle Tennessee, huge ratings, TV wise for Alabama games. It balanced out this fall for some reason. Also, no, no idea why. No, a lot more tease, a lot more orange started showing up again, but it's still a Tennessee market. When they're good, it's people are really into it. They have to be good though. But like Nashville's, I love this city. It's my home, but Nashville will leave the party pretty quickly if it gets dull. So you got to keep people. I remember being here, I don't know, 8 9, ten years ago, whatever it was and they were winning. Oh yeah. And everybody was a predator, Sam. Well, the cup run in 1617 changed everything. It brought in the common fan, the casual people, Broadway was insane down here. And even since then, the city has changed it, but there's about 14 more buildings downtown just since 2017,
Jason Zucker scores in OT, Penguins beat Lightning 5-4
"The Pittsburgh penguins extended their win streak to four straight with a 5 four overtime win against the Tampa Bay lightning. Jason Zucker got the game winner as well as Pittsburgh's third goal. We're just trying to play together. You know, trying to help each other out. You know, we know that we're going to be battle and we know we're going to end up on our heels and different parts of the game and jar stepped up tonight. I thought our pKa was unbelievable tonight against one of the best power plays in the league. Jeff petry also had two tallies for the penguins drew O'Connor scored Pittsburgh's other goal, Braden point finished with a goal and one assist for Tampa Bay, same for Steven stamkos. The lightning have now lost three straight. Tampa
Young's 20 help Maryland surge past No. 3 Purdue 68-54
"Jameer young led three players in double figures with 20 points as Maryland upset number three Purdue, 68 54, handing the boilermakers their third loss in four games, a technical fall against mason Gillis with per new up 8 helped spark the terps to a 29 to four run in the second half. Young says that's when the game turned around. Just trying to limit them to one shot, get out and transition. That's when we play our best balls, so the crowd was into it. They was bringing energy and, you know, we was taking good shots on our end. Braden Smith and Zach giddy had 18 each. The turps out rebounded the boilermakers 35 to 23. Craig heist college park Maryland
Vasilevskiy ends 84-game shutout drought, Lightning beat Avs
"Andrei vasilevskiy posted his first shout out to the season at Brandon Hagel collected three points in the lightning's 5 zero thumping of the avalanche. It was the 29th career whitewash for vasilevsky and his first in 85 regular season starts. Ego had two goals that an assist and a rematch of last year's Stanley Cup final, which was won by Colorado. I think tonight we played very well. It's obviously a matchup that we circle at the beginning of the year with everything that happened last year, and for us now going on the road that was a huge game for us to kind of start and get more into going the right way. Corey Perry, Braden point, Mikhail sergachev also scored in Tampa Bay's first one in three games. I'm Dave fairy.
Braden Smith leads No. 1 Purdue past Iowa 87-73
"Purdue freshman brayden Smith played more like an upper classman in the boilermakers 87 to 73 win over Iowa. Smith had 24 points as number one Purdue improved to 23 and two. Shots just fell and kind of just play how it went. So with shots being able to fall, obviously confidence goes up a little bit. So that's what was happening. So it just felt good to make a couple shots. The boilermakers saw a 21 point second half lead cut to 6 with 6 minutes left, but it was all Purdue after that. Chris Murray had 24 points to lead the 15 to 9 hawkeyes. Tom mccabe, West Lafayette, Indiana
"braden" Discussed on VUX World
"Gone down for a different more recently in their dedication news. I think we've certainly seen demand kind of fall off there. But for Alexa, we've continued to see continue to see it. I think what's changed though is demand for assistance across chatbots, bringing in all used traditional call center stacks, which might actually not have had an before. They're sort of more legacy on prem technologies as they move to the cloud. Interesting, more of an assistant architecture. I think it's just a demand for those of skyrocket. It's a relative, you know, relatively speaking, it's not that demand for Alexa has shrunk it's just demand for chat bots and assistance in general as just a skyrocket. And so that's been the majority of our business over the past two years now really is trying to become. The company we essentially want to build is almost like the Adobe suite for conversational AI teams. So having different products that enable different elements of the workflow, but together can be one conjoined unified workflow. All the way from design prototyping, managing of content management and all these kinds of things. Either offering integrations, working with partners, having as many open APIs as we can, so people can build their own custom workflows with whatever tool stack they want to use. Because I think the only certainty of where this industry is going is that things are going to continue to change rapidly. That's really the only certainty we have, right? And so we're just trying to make sure that we build our tooling in such a way where people have the flexibility. So as they want to plug and play different tax again, that's probably the cornerstone of how we're building building our platform is we want it to be open to work with any text to speech, any NLU platform, any underlying dialog manager. The only thing we're trying to do really, really well is enable that collaborative design led workflow and then the underlying technology can be whatever you want. And so a bit of a tangential answer to the initial question.
"braden" Discussed on VUX World
"That is the UX dot world slash C or G and I G, Y okay, let's do this. So without further ado, let's welcome a longtime friend of the shore. Mister Britain ream of voice raw Brandon. Now then. Hello hello. Good to see you. Hey, nice to see you again, my friend. How's things? Thanks for good. Things are good. I'm actually in New York just visiting some customers this week. So it's kind of nice, you know, when I will never turn down an in person meeting invite. So for anyone watching, if you're in a cool location, let's go for coffee somewhere. Definitely definitely. Yeah, yeah. I hope we're going to get to ourselves to New York ourselves in October when we do this event, maybe swinging by New York, you know what I mean? Spend a few days there, which would be quite nice. Never been a new year before. Beautiful. I mean, it's been, oh my gosh, like what, two years since we did this podcast last. I think at that time, we were mostly Alexa Google. I think we were just starting to get into conversation design. So it's a much needed refresh video today, so definitely. And we've got quite a few people tuning in actually, which I suppose is nice because this is also happening at the same time as Alexa live, which you pointed out when your email to me the other day. And so it's interesting because it seems to me, and I don't know what you think about this, but it seems to me that the attention has drifted from voice assistant platforms and moved more into the higher level NLU and LP category, voice assistance has been part of it, but obviously there's a lot of traction in the IVR space, which is what we spend a lot of time doing on the chat space, our messaging, all these different channels and environments that can utilize this technology. It all just seems to be blowing up at the minute. Yeah. Yeah, I mean, I think as an industry we've defined what we want to build towards now, which is every company sort of having its own its own assistant. I think Brett console did a really good job. Sort of highlighting this almost two years ago now. But we had a walled garden approach with a lot of the voice assistance. So think of the Alexa as the googles. As almost like a netscape. It was a browser of what you could browse domains that had to be built within its walled garden. But now we're moving into an approach where companies want to actually own the tech stack. They want to own the data. They want to own the customer relationship. And they want to actually build their own assistant that can work across all channels. And so it's a pretty exciting future. We talk about this internally as we're trying to help enable the world of a billion assistance, right? Where every brand will have its own assistant, eventually I think we get into personal assistance, I can intermediate and do larger, more complex complex tasks, but actually working with the.
How Manufacturing Platform Raven Found Opportunity for Growth in the Medical Field
"What does the future look like for raven the product and for your team. I think it's it's really bright. It's a really exciting time for answering right now. So we've we've grown about thirty percent in terms of the team in the last six months and we're delivering features functionality improvements at faster pace than we've ever done before the potential is enormous in this industry. I'm forget the number here at the time. There are hundreds of thousands of manufacturers in the world. it's a it's a huge part of the economy and there is huge potential for improvement so we see a lot of a lot of great things happening in addition in story here at the beginning of the pandemic one of our first employees or i think he was technically the first employee gave us a call and he's now a surgeon and he said hey. I think i think raven could be useful in our operating rooms at our hospital to help us with our cova productions to keep our our staff safe and and also to just help us be more productive over you know. Use this sort of scarce resources more effectively. And so we've been you know looking at how that works. And we now have of a variation of our product installed as an operating which is which is pretty. We do see you know at its core. We're trying to help people use their time. Better use their resources better. And that's not just manufacturing but it's also a lot of other areas manufacturing stuff is is super exciting. That's our core but we do see this potential for growth expansion into these other other areas as well
Raven Founder Believes Fixing Mistakes is a Team Effort
"Let's flip the script a little bit. Tell me about a mistake. Made and how you and your team responded to so this is one of these things that specifically talk about. We've all made mistakes of varying degrees. And i can think of of cases where we've accidentally deployed wrong pieces of software or deleting a node where we shouldn't wear we intended to upgrade a single pod or or something like that the amazing thing whenever we have sort of one of these issues that comes up. I love a jumping in whether it's a discord room or slack all or whatever or you know back when we could over the shoulder. The team is just vicgory focused on solving the problem. In a non-judgmental way. And then we we. After the fact we do failure causes inquiry where we go through what happened and in and realizing we're part of system here in it's not the individuals it is the system. You know the when you when you mess up you know your your heart starts beating you feel that that anxiety but you've got a whole bunch of people who were there with you trying to help the system backup and then support you as we all learned from it and try to make the system more robust so that that doesn't happen again unlike what she said there about the system. We're a system. Were not an individual. So how do we fix the system. That's a good. It's a good way to scope it when you have to have hard conversations or no retro on an
Raven Founder Braden Stenning on Achieving Flexibility for Optimal Scalability
"What's flip the scale ability. Did you build scale efficiently from day one or were you fighting this as you grew and gained traction one of the things that we wanted to prove that we had a product that was worth building and from a technology perspective. I think it's very large stack that we try to. We try to do the end. End being data collection data cleaning to get to the truth of the analytics to the insights of the feedback to operators. And so there's just a very large technology stack so when we build our technology we would try to pick an architecture that is extendable extensible and it's flexible and we wouldn't necessarily build all the scale ability to start but even try to leave ourselves the options from a product perspective when we built our product. We did think about scale ability absolutely from the beginning to be able to help operators at an individual process we wanted to be able to help the corporate executives overseeing one hundred plants that mike have hundreds of production processes and so thinking about that product skill was was something that we do did from the beginning thing about the technical scale with something that we considered but didn't necessarily always implement we'd hit the limits and and that was through continuously monitoring the performance of the system. We would go back and we would adjust for our particular case. It's an appropriate way to to build it small team looking to move quickly deliver quickly and it's about that speed of providing value to our customers so it sounds like you started out building in such a way that you could prove out the product but you architect it to where you could scale easier later on down the line that right. Yeah and we're continuing to do that. You know so. We've made some technology changes and we're continuing to make those changes in the last couple of years here. We've we've really found a lot of great product market where our customers are they want. They want things they want them. They want them quickly and building in that flexibility and the technology is something that we you know. We now architect in that flexibility in that quick speed of being able to adapt and taylor to them. You know we're building a product so we want to build the product. That's it's it's flexible because that's what we need in this market
"braden" Discussed on Doin it! with Danny and Jenny
"Yeah, he's so good in this so great and it's a completely different show this season but I still, I still, I watch, you know, it's got me, I'm still hooked in it, and then I talked about last week, have you seen yet the mysterious Benedict Society page Not I'd say I saw it on but I did see it on on, on visiting. It's okay. I'm going to, I'm going to pressure you for next week to watch it because I I watched the third episode. I really, really love the show took. I Still Loving Loki, by the way, I walk he's fucking fantastic like just love it, it's great. Yeah, Loki is fantastic but that mysterious Benedict Society, Tony I was so good at it. Okay, I'm now okay, this is a good recommendation, relax, I love Tony Hill and any I feel like there's nothing Tony Hale everything. I love right now like the house broken show. The cartoon he does, like re voices on that web. I want to give that a shout out cuz that's Jen and Gabi who created that and and it's so good to have spoken. How may I show about? Like a group of pets who are in like group therapy? Yeah. And it's it's super funny. You can watch it on a Hulu. It's a it's a fox show at least across Voices, the kind of the main dog Clea Duvall. Some of the creators to really good. But Danny would have you been watching? I mean, I mean those two high-scoring? Mm, she gives my sport. I shoot. I'm going to watch some more later. Who's your favorite? Pornstar? Currently might ask her all girl anymore. I can't really name the women anymore but I'm like, you know, I like the old Oldies. But Betty White is still your favorite. Yeah, really Betty. White is Faith late but a lot of episodes and it's Moody and weird and Icelandic drama both call called kotla. And it's like there's a it's so beautiful but it's like there's a there's a volcano on Iceland and it's very long. Low but it's like there's been a volcano and out of the dirt. This like sort of this woman sort of emerges in his lost. And and and it's what's affecting this town, I haven't sussed out the laundry or what it is, what they called again. I'm looking this up. K, a t l a. Okay, now it's Iceland. I don't know. Like there's something about these foreign shows on Netflix. I love because I don't recognize any of the actors and I can really lose myself in look story. Yeah, it just I don't, I don't, you don't, let's go like oh, I just saw her on page six. Exactly like throwing up on her ex-boyfriend. You're like, oh no. I don't have any idea who this is. This is kind of well and he'll, I always Delight. I hope we get to see you in person. Are you? You're in La choya? I am. Yeah. Yeah. We should we should hang out soon..
"braden" Discussed on Doin it! with Danny and Jenny
"So I just started sending in cells constantly and like, literally the first song songs, I sent in every single one of them. Played immediately. And no, it was amazing. Like, I remember listening and Howard, like just hearing his voice go and another one by Library girl. Like he just made it. So like matter of fact, it's like here, we all, you know, another top hit from Glenn Frey, everybody know he just said it. So matter-of-fact, I was later. He shut the fuk up Eli's juice and killing it every day and there's no no matter what you don't. Anything that ever happens to be in, you know, my career that will never be anything that tops that because he's say no way Jose. Reach of it. It's so personal with Howard like it just feels like like everything. Like I like it's I always tell Gary this but like you know I've had you know, could have a show that aired on Project or whatever. If Howard even mentions me in passing in like a thirty-second clip, I hear more about that than I do about anything else. Yeah, I I get it. I totally get it, you know? I mean obviously you've had your game even such a huge success. Huge success with Howard? I mean, huge success, but how you want to cancel, you want to get your enemies? Where are they, where do you keep them from there to have them in the my own background and listen to me? I mean, but is it? Yeah, Melissa took a bit but when Howard mentions you that's where I'm like, I want to reach out to Danny. Of course, cognizant of years, you're one of those people who like will tell me that the guy is really good job. On it, but there's like a bunch of people just like, it's really fun. And and so I understand like to have it as much as you're having it in to have that reach. Are you still like would you like I know in the beginning you must have been writing a couple of day for him, right? Yeah. Yeah. For sure. Are you still at that level right now? Are you have you have you tapered off or do you like when I'm not working? When I'm not working, I'm totally into it. But like, I mean like, you know, I basically, I mean, just long story short, I have a deal with the show now, because they pay me and I, I basically just got to produce like I basically after. You know, two or three songs a month but like, normally I go way over that because I, it's not about, it's not about like, oh, I have to get paid. It's about the question. I was not about the money. I wanted to get married about the money for that..
"braden" Discussed on Doin it! with Danny and Jenny
"Write a song about how no one will ever see her this song because everyone's going to skip the credits. And so, I recorded it, I wrote it recorded, they loved it. And then, I mean, you know, it's me just being like, hey, I gotta push myself off. I gotta do more. I'm like, hey, what if this is like, an evolving story about this guy? Who's, you know, like he's like convinced like no one's hearing him so he he can like just say whatever you want. You can just like kind of like tell this story about like, hey not only his Noble listing. So I'm going to say what I want to say and they're like, we'd love it. Just run with it. So I wrote a few more versions there. That's great choice in that first season. I did a couple different versions that as, you know, like her for like two or three episodes would have this song, the next two, or three episodes of this song. And then, and then Thursday through got approved. And I'm like, guys, what if I do a different song for every episode because I'm going to get paid for every one of them of course. So they were like, we love it, we love him. I had one where my daughter was singing and the whole premise was, I was coercing her into doing it for money. It was exploiting her, I had one. Just it was every every idea I could think of became, you know, this could have gone Four Seasons seasons of Seasons, but but Netflix decided, hey, we don't want that. I actually remember wrestling with the the interface while watching it because you had to click pretty quick immediately immediately. And then you'd miss it and then, like, I go back and I didn't get to the same point, you don't work either. See the end credits, sometimes you have to start from the beginning of the episode and then Southward all the, yes, I did that with other shows, like, you know, when there's a voice, unlike say big melt, like, who was that? I am so, there's a voice. I'm like, yeah, they they, they made me really make it hard to see those credits, right? And any time I ever want to see credits, like sometimes I'm like, oh, this is good. Why do I have to hear this whole theme song? At the end again and it's going to go and we're going to see the locations. You are at like when you're seeing the credit and and then when you want to see it, it's like boom, skip it. You're done one of the next show you like, actually I think one of my friends was in that I'm pretty sure I saw her. I just went to see it. Maybe imagine. Yeah, that's like they make it so hard for people's parents to be proud of them and whatever else and you had some funny recurring bits on that show too. Like I don't know what you know what your role is, but I love the guy who was it was the pizza box, man? Remember like he would come out and he was like he was the guy from the pizza box..
"braden" Discussed on Doin it! with Danny and Jenny
"Danny, I'll let you, you can show your, I mean, I don't know how much, you know about you, but I was, I know, I know all about your certain history wage. It was roommates with Baba Booey. So I can. I remember that moment? Because I knew you prior, I think, For those who don't know, Eli is a major contributor of of songs about Robin Quivers. Hits that are played and how many would you say you have all my life? I mean, I've done, definitely like at least five hundred songs about Robin, you know, since the pandemic the show has, they don't do the new segment anymore. Oh, so it's like, I do all kinds of songs, a lot about Bowie songs. A lot of songs about various Colours Bobo king of all blacks, a lot of songs about various staffers, you know, Benji, Sal may I love I love doing a chance to but but yeah, I mean I I continue to worship Robbins tits but there's not as much of a there's not there's not an opening on the show for four songs about her kids more. So yeah, I think it's important for an artist to grow. So you know what I think it is a natural like just as you know, when Bob Dylan went electric and you moved from Robin to Bobo. Yep. Exactly. I mean these are but Robins titties need to be celebrated, always think. It's fair to take this away from the life that they lived. There are other contributors of of music. Is there a competition of camaraderie? Are you a fraternity of like, who are who? Who are the, you know, who? When you shout out to his Inspirations or people who make you a better song parodist in that world? Well, among among the song parody people for the Howard Stern Show, I do have a strong Alliance. Some little Mikey. Yes I am and and wage there was there was one point in the show's history. Probably you know, eight years ago up until maybe three years ago where the main three people was me a little Mikey and psyche and life cycle was blind and I was not a fan of Psych. I went on the show a couple of times with psych my disgust how much you hate blind people. This is, you know, I'm Not Dead. The Olde blind people but most Marlee Matlin definitely on my shit list when she just can't hear she just can't hear. Oh right, okay. Well it really is everybody with disabilities that I'm do. I get that? I get that. I mean it's a white as a straight white male. I'm kind of like, you know, all of you just go away, you know, it's like, you know, whatever. Modelling someone standing up for human straight wage is a chance. It's, I mean, finally sort out your first. When are you coming? Stevie Wonder Stevie Wonder wrote a couple of good songs. Let's be honest. I mean I'm sorry. Okay he's black fine..
The Role Shift of Voice Assistants to Search Engine-Like Voice Aides With Braden Ream
"Would hope we start to see more intent based queries for the actual general assistance Googles of the world where are no longer asking acknowledged based queries like hey hotels april tower. It's i'm hungry right. It's more of an intention and we start to see that consumer behavior shifts. There's a bit of a chicken and egg Around this but let's just accent assume or a perfect world for a second when you see that consumer may shift words. I'm hungry or you know. Where's the nearest restaurant is another way to say i'm hungry right. You're implicitly saying that you're going to start to see these assistance act more like search engines which is really exciting. Because i think that's when You know we can start to viewer assistance. The reason why they're called assistance right. They're meant to be helpful mill to help us in our everyday tasks not just voice you know. Search of google acceptable often with a lot of assistance are for most folks. It's whether alarm clocks ability to search for information We want to be able to Access manipulating engage with information right I think we'll definitely see that in the next five years. Ten years is like without with extreme certainty Two years is a little bit. you know. It's plausible and then in five years. I i feel pretty confident that you'll start to see the voice assistance that we all know in us today really become search engines and the allowing access the world around it access information and access services around us really with the use of a personal assistant concierge that we have in our home.
Flamingos Can Be Picky About Company
"Spend some time watching. Swimming goes and you might think that goes on tiny heads but these elegant avian actually lead complex social lives. Each bird has certain other individuals it prefers to spend time with and others that avoids in other words bingos have friends the wildfowl and Wetlands Trust the WW T. manages a number of wetlands in the UK some of which have communities of captive aquatic birds including bingos. They didn't quite know whether they could just take it for Bingo. I to the environment is ticket new flock and it would be fine will. Should they care more about the social choices at the birds? Were making Paul rose a psychologist at the University of Exeter's Center for Research. In animal behaviour for five years rose in his team observed the daily goings on of five. The world's six different flamingo species housed at the WW T. slimbridge wetlands center in Gloucestershire. The five species were the Chilean and Andean American James and Lesser Flamingos the partnerships that we say between bad saw Don Braden. The badge are choosing Haizhu associate with male female pairs spend time together so do same sex pairs and even groups of three or four and those relationships can last for many years. The findings are in the Journal behavioral processes so there was some flamingos in the group. That really didn't care who the caught was for that day they would flittering and they would have many different relationships in many different birds. And of course there were other flamingos. That were less social butterfly and more lone wolf but even they had a few close friends. Those beds that at least Gary s had more investment in a small number of sexual bonds. Living that they knew really well. Rose thinks that for Mingas may have evolved their social lives due to their wetland habitats in which the resources they need are concentrated in a small area. The Sexual Organization is layered on top of this need to be in one environment. So you have to be gregarious. If Flamingos know that partners if they know that the six birds off friendly and they get on with them they can then waste less energy scrapping and squabbling with other birds or they didn't get on with by spending time with their friends. Flamingos can more efficiently director limited time and energy to active these. Like foraging and mating if I want to get one standard one in pre my feathers. Garin stand on my leg and prayed my feathers next to Fiona because I get on with her all the next of frank because I haven. We'll just scrap if all this sounds familiar while we all know. Feel now as well as a friend.
Washington, DC - Braden Holtby taking it easy in quarantine, even with uncertain contract situation
"The NHL season remains on ice metaphorically instead of literally how is capitals goaltender Braden Holtby passing the time do much right now Mr in the summer and takes a few months off completely anyways is trying to get my mind completely away from everything so is it's gonna be tough once there's kind of a game plan and play whatever we have a bit of time we're going to we're gonna come back hope you had a goals against average of three point one one the highest of his career when the season was suspended
Redskins put franchise tag on guard Brandon Scherff
"Even the Washington amid widespread Redskins a place postponements the franchise sub tag sporting on three events time worldwide Pro Bowl guard Japan Braden says sheriff the twenty the twenty two thousand summer fifteen Olympics first in Tokyo round selection will had go been in on talks as with planned the club despite on a long the corona term deal virus outbreak instead the Redskins Japanese ensure prime he would minister not reach Shinzo free agency Albay said Saturday Washington the Tokyo use the regular games will franchise still take tag place meaning in sharp July could still without talk with basing other teams Japan will the host Redskins the Olympics would have the right quote to match without any problem offer or as receive planned two first U. round S. draft president picks as Donald compensation Trump suggested earlier they could in still the week negotiate that Japan a long term should deal consider until postponing July fifteenth the Olympics the franchise as of Friday tag property Japan had Wyman over fourteen is expected hundred to be confirmed worth upward cases of fifteen of the covert million ninety dollars virus this season and around thirty Michael deaths Reaves I'm guessing Coolbaugh
Ovechkin nets pair of goals as Capitals beat Wild 4-3
"You know is a pretty good night for the capitals and wizards they were winners on the road tonight to rebound from road losses the caps beat the wild four three in Minnesota they got a pair of goals from Alex a veteran and a big third period score from Tom Wilson and stood as the game winner Richard Pontic adding the other capitals goalie Braden Holtby thirty seven saves as the caps Snapple four game road losing streak they up their division lead over Philadelphia to three points entering their head to head matchup on
Durham, Jackson-Davis lead Indiana past No. 9 Penn State
"T. in J. a game Oshie of extreme scored the go runs ahead goal Indiana snapping a three had all the one tie that midway mattered through most the third in a period sixty eight as sixty the capital win snapped over a four number game nine losing Penn streak state beating the penguins the Nittany five lions to three were down by the nineteen take over first late place in in the the first metropolitan half division but led by six the capital ten minutes scored later three goals in Indiana the first eleven answered with minutes the thirteen of the third to nothing it spurt was the penguins and held the lead third down straight the stretch loss Carl outdoor Haglund led and a the pair Hoosiers of goals with fourteen including an points empty netter as Indiana late improves the penguins its got big tallies ten record from Patric day in an Hornqvist eight Sidney after Crosby winning eight and Marcus in a row Pedersen Penn state has dropped Braden its Holtby last made two thirty three Lamar saves Stevens in the win led the Nittany Greg lions heist with Washington twenty nine points Tom McKay Bloomington Indiana
Capitals snap skid, beat Penguins to vault into first place
"T. J. Oshie scored the go ahead goal snapping a three all tie midway through the third period as the capital snapped a four game losing streak beating the penguins five to three the take over first place in the metropolitan division the capital scored three goals in the first eleven minutes of the third it was the penguins third straight loss Carl Haglund and a pair of goals including an empty netter late the penguins got tallies from Patric Hornqvist Sidney Crosby and Marcus Pedersen Braden Holtby made thirty three saves in the win Greg heist Washington