Creating super-human voice assistants with Nikola Mrki, CEO of PolyAI

VUX World


Apple for a while and then you win and create polly ai. What was it specifically the kind of took in that direction. What was it that you enormous in wasn't working either necessarily broadly speaking invoice technology. What was it that you saw wasn't quite right. In what problem did you think you could solve by by creating paula. I absolutely so What i think the challenger creating a horse assisted. Level thrive people through any of these media kardashian mobile phone speaker. Second thing it's different. Hotter product has many uses. The finish quality and other food is very different. People inside cleanable the to clear those the ottawa Negative while we're kind of like you know the future Know say amazon charged enough to have On every two will not when you aggregate fact their fifteen thousand equal working at alexa do actually as long as someone who worked on this before it was as huge as opulent as it is today. You can't help but things have not achieved back much when you know like the few scenarios that he knew any of these technology despite music you know setting larssen timers asking you questions when playing board games navigation of exiting but overall. It's really not that overwhelming. There wasn't hope that this would be the next thing after mobile really. it's kind of growing. It's growing the read. The say laptops grew up in the ninety s and the group right in compound. Interest is miraculous dion here with a laptop. But it's definitely not exploded in the way to save the iphone. Did even though. I think many of these companies that would now you hear these very articulate narrative Directly to shop is enter. Be really materialized when we have a are in vr blindly. Those are all things that will build into insomnia clear where He conditioned from the ecology by co founders through our senior research team. We were all brought up by. Crm's steve is less cambridge professor. Almost one of the guy had species mission for years. Was the leading person. Pushing hidden markov models. We wanted only way. You'd bisbee shirt mission until these learning got an and really as a matter of ernst and they were more efficient methods previously Westchester kangaroos he sold three counties As he you know supple the see that since two thousand six hundred billion dollars bermudian six Missionaries are twenty thirty years. It is impossible for measuring that you can have a sophisticated conversation with us. But see sorta developing formalism the models these errors is waste recovered specials severe to really bring converse in to the understanding. So that you're able to really elevates something we really awesome A few people in the company that there are similar problems leader in the battle management decided save depending on. Both the of the user might have met Sheet metal wall seats available shop inventory or reservation book and so on. Yeah i mean when when when you think about what has been built in where you go. Silence made sure to make progress. The most isolated followed that as scientists Both saw its customers. There's a lot of data. There's a genuine genuine. Consumer eve right at All these fools and interestingly i land look you go into a large People customer experience with people bought line of the cio's and they all really end up on with your yard also. And if you haven't done this before we ever all happened so you are brighter citizens. We were the city when we had a lot of success for us. Few years is really a shame of systems that are incredible airless. Some of the most sophisticated thought An employee narrow the bird. They're not here and talk to you about anything we can talk to you but meeting life but you know you want to book a table and you really or schedule. Data's choosier through the evening. Not after learning must really complex stuff. The the more by the customer service is are actually able to evaluate the systems breath. And how he florida track assistant done and that. The second problem is blake all frustrated. Leave users fears to Speaking we choose our battles but out we're able to perform the most human agents quite often we beyond to be probably won't really into application just until how do mean by outperform humans in in what respect to outperform. So that's your accomplice crush on the on the one hand just reaching characteristics. Aw a support. Asians or human didn't make them popular the audio voice. Speeders your other understand. The user meets all jeering answer. It didn't constraints where it is more narrow occasions where eagles Now when he reforming hugh. If i stayed in my last elections naza were. Wade's speakers Are you paying and quickly. Say that my vote is was that might or bamberg for the for scherzer. God's actually slow surprise a matter of fact but now we're able to mav performed see humans wanting to actual research the school of cool still understand really talk to the salsa just beating the baseline but there are very few most sectors so well we've been there for long enough to receive a very product where they're hundred about it like sizes of all those questions even though the answer because it's more bright they how the answers. They're connected to social system. The answer changes you know media the people because they straightaway. Don't get tired whatever language you need them to use Something's the edged. Change the So there are many hours which museums are falling like. We speak a few languages. You know we don't know the answers. We will startle both codes and in many of these many of these problems we have transactional but really currently functional assistance in place authors.

Coming up next