New York Times, Martin Ford, Ray Ray Kurzweil discussed on OC Talk Radio

OC Talk Radio
|

Automatic TRANSCRIPT

Know we're back with our best. Selling off their New York Times bestselling author and the author of the new book architects of Intelligence Martin Ford Martin Welcome back and <hes> let's talk about one of the things that <hes> for my listeners their business leaders and they're looking for tools advance their company as other technology side. There's always more terms dealer work and in a is no no shortage of new terms. I like in your book. You've put in not just a definition attorneys but you've described <hes> and tried to walk people through them so they can understand what are some of the important term zany understand <hes> with the well there there are a number of terms that are very important in artificial intelligence and you do need to kind of have some understanding of if you really going to have a good conceptual understanding of the field so for example there are different ways in which machines learned. There's what's called supervised learning which means basically providing <hes> many many any examples maybe millions of examples of data to a learning algorithm and then training it on that basis you an example would be training system to recognize images of animals so you would give millions of photographs each of those photographs of the label right. This is the dog. This is a cat. This is a cow and it's just been looking at millions of those would relearn how you do it so that's what's called supervised. Learning unsupervised burning is is much more difficult. It's more like what people do we learn without having all of those examples downpours right <hes> <hes> child calmer n- what the animals are without certainly being given millions of examples. That's actually one of the most intensive research areas in artificial intelligence is is figuring out how to teach machines to learn in an unsupervised way I <hes> then there's deep learning which is all about neural networks the hottest thing that's happening in artificial intelligence right now so it is important to have some understanding of what these terms are and what I did in the book is right a very brief. It's really only a few pages is that Kinda introduces these terms hopefully in a relatively painless way so he can quickly get a sense of what they are and what they mean and then after that you can delve into the interviews where you actually talk to your she talking to these he's been caught. Minds and artificial intelligence and cars will come up and of course those interviews and I'm and in many cases I actually asked these people to explain those terms and what you're getting here is the chance to talk in some cases to the person that actually invented this contract. You I think the guy that actually created this so it's a little bit like <hes> I imagine you know talking to Einstein and having him explain relativity relativity to you or something like that right these are really people almost at that level in terms of you know how prominent they are and they've invented a lot of this stuff and they actually explained some of these concepts to you and so she gets a primer and in the beginning where you can become more or less familiar with these terms in on delve into it <hes> in a little bit more depth. I think it's a pretty painless way way to do it. It's more interesting maybe than trying to read a textbook or something right. It would be very very dry here. You're actually engaging in a conversation and people are talking about how they discovered these principles and how are you going to be applied on wider import so I think it's really really quite a good resource for anyone. That really wants Kinda the inside track on a I you know this is a chance to get inside the minds of the people that are building this technology the most famous people in the world and find out what they really. We think about the future and journey anyone who is an investor for example unless to invest in a are wants to know how hey I is going to impact their investments. I mean this is I think a pretty good opportunity. I wouldn't pass it out myself your lot about going on the stock market and how much the machines are making decisions on buying and selling of stocks and for funds and things like data. There's quantitative funds rely on Algorithms is are they using a they're absolutely yeah. There's some companies like blackrock for example. <hes> didn't invest in his very very heavily have very high expertise Ortiz. They are building you know intelligent algorithms to trade stocks and in fact as as you may know most trading on the stock market is now out rhythmic spot. He's not performed by by Humid's and these algorithms are just incredibly capable of course they're incredibly fast so they make trading decisions. You know just miniscule fractions of a second long before anyone can even make decision to to reach or for button to push you know it's all finished y'all Graham is done on and and these algorithms can just as an example they can tap into machine readable news sources so companies like Bloomberg. They actually published news. That's in a format intended for machine story. Not People and machines can read this news much faster than any human being make sense of it and then trade on it. <hes> you know in in miniscule fractions of the second he's algorithms also they badly chartered. They do things like they place decoy trades and then they withdraw trades in order to try to bake out the competing algorithms or or human traders <hes> sort of already very sophisticated. Some people have said that in some cases the algorithms are really going beyond and what they're human ventures even fully understand and of course that does lead to risks right. I mean we saw a few years ago. The flash crash for example where some people think what happened there is a budget algorithms. There was kind of a herd instinct right where they all started did showing at once and there is a danger of that kind of thing all these trading rooms decide all at once to sell and that's that's not going to be good <hes> so that's kind of their team right. They get stuck you know he would but when the machines are actually gaining each other right. They're trying to figure okay. I think this is other is other someone else doing this so we're going to reactor their gaming it <unk> out and so they're they're gaming it out on a faster faster timeframe right. I mean you know it's happening. In in tiny fractions of a second you know I mean at a speed that no human being could possibly comprehend so so they interpreted that kind of competition timeframe. There's no there's no pace for a human trader at all they having to remember once there was a they had to pull the plug on one one system sure it's happened more than once but <hes> out of the House the human keep keep tabs on this. That's the machines are trading in billions of dollars. They monitor it. I mean they they see what's happening and they can monitor it a range but they certainly don't have control of it you know at the second by second level. It's much more of a macro view. <hes> and many of these systems are used more for short-term trading what you would fall day trading or <hes> and human beings probably still at these currently have more of a role in long-term by trading decisions but <hes> in terms of you know the timeframe of of seconds. There really isn't any way hey a human being control in control it would that kind of granularity I mean but <hes> so it's much more a case of stepping back sure so we are talking about the we talked about the economic disruption. We're talking about. There's there's a pass towards the human level artificial general deligence the A._G._I.. That's that's occurring now but it's still not quite there yet. They're against for example would be from was it. <hes> the movie talk two thousand one and our nine thousand. Is that that sort of human level of intelligence. That's what that's what we would call human level. Hey I or or A._G._I.. Artificial General Intelligence and examples would be like you said two thousand one space Odyssey <hes> the computer in the starship enterprise and star trek or or commanded data <hes> in the Matrix movies the agent agent spin it those were all machines that have human level intelligence <hes> and the probably said they exist only in science fiction. We don't have anything like that. In the real world you do have little very very rudimentary things like Alexa right you can speak to on some level but there's nothing close to human level intelligence there and in fact we're not most people think we're not very close to that and yet this has been sort of the goal of the build artificial intelligence twice since the very beginning when when Alan Turing wrote his first paper on A._I.. In nineteen fifty invented the turing test which is basically a cast or true intelligence that says if you can have a conversation conversation with a machine in a weighted it can deceive you so you can't tell it it's a machine another person down. We'll say okay. That's probably you know an intelligent machine and that was way back in nineteen fifty that was he was already thinking about that or we if you are anywhere close to that so one of the most interesting set of conversations in the book was everyone about what does it take to get there and how long might it be and there's lots of variance there. I mean <hes> show the more ambitious people like Ray Ray Kurzweil. A futurist and many people have heard of the guy that talks a lot about the singularity for example anything <hes>. It's just about a decade away by twenty twenty-nine. We'll have human level. I but that's a very aggressive prediction. Most people people think no way that's not going to happen. <hes> one of the other people I interviewed Rodney Book Brooks who is the <hes> Co founder of I Robot Corporation the company that makes the room butts. <hes> said it could be nearly two hundred years in the future two hundred years before the machine knows not to do the whole house just right one two hundred years before we have human level I chewed thinking yeah down with and have a conversation right and and it's interesting Christie not but to take that example to have a truly intelligent vacuum robot that that really understands everything you would want to know about vacuuming your house. You almost need to get to that pretty close to that level right to have that kind of intuition and the judgement so <hes> you know it it it could be quite a quite a ways in the future but in the meantime we've got specialized artificial intelligence right. We bought what we've got now machines that do specialized things that can power a self driving in car that can look at medical image and decided that cancer or not it can crank out a new story whereas nets but that's fine it. They're not they're not looking to do all kinds of things are not class the individual machines czar either playing the Chinese game go or their vacuuming vacuuming your your house. They're they're not doing everything right. That's right. That's an important distinction so they're doing specific things but the other thing the not forget. It's like I said there's going to be an explosion in the specialized things so there's not going to be one system that can do everything but there's GonNa be a million specialized ones they can do you know anything you want basically right that that at least anything that is under mentally benchley routine and repetitive in nature so in terms of the impact on the job market the thing to keep in mind as a lot of people do fairly specialized routine repetitive things right and you don't need human level artificial intelligence to do that. You just need specialized artificial intelligence to do a lot of that and there's going to be an explosion in those applications in the recent explosion is that related to how much data's now available and how if oh can Bernau machines can now access specialized data right so there are three things that really have made this revolution possible one is there's some real breakthroughs in terms of the technology things like we mentioned Jeff Hinton earlier right. He's one of the guys that really made this happen..

Coming up next