welcome to another episode of Jasmine call Travelers on Perris and simply mind bogglers my name is Tolson Jacoby and I'm your house this episode of judgment call is funds at the mighty Travelers premium full disclosure this is my business and if they have a 450000 afros everyday to give you the best deals in economy premium economy business and first class we also make recommendations for four and five star hotels all over the planet when they are much cheaper than they usually are a thousand subscribers have saved more than 95% the airfare tickets and a flaw in the business class lie-flat transcontinental using audio in case you didn't knowAmericans Europeans and many other nationalities cannot travel to more than 80 destinations again to give it a shot and try of my teacher looks premium for free for 30 days today you can sign up at my travels. Com MTP of everyone who's trouble with all the characters go to MTP for you that's just five characters MTP for you.com I'm very excited today to have Steve Schwartz as my guest on the Jack McCall podcast Steve is an author investor and serial entrepreneur and Dead 1981 Steve co-founded the first AI companies call cognitive systems and Steve has taken up a lot of interest in to REI again after putting the top of a little bit of brother in the nineties Steve also wrote a book free ebook artificial intelligence 101 which is available on his website bi prospective stuff, and his upcoming book is evil robots kill our computers and other myths the truth about Ai and the future of humanity I see if it's great to have you how are you good high tourist and thanks for having me that's a big topic that you have for your book it is yeah I'm excited about it coming out what what is the main theme of the book you know I can I can see that from the title of his leave it sounds like you're you have a different view of the eye than most people have I said most people because that's his right now is that the AI is this enormous success of the last couple of years and then we will see the same success some kind of oil changing success to be seeing the last five years we will see more of this in the next 5 to 20 years you see the slightly different correct a little bit different yeah hey I has clearly made great strides from an engineering perspective Siri answers are questions Google translate helps us talk to taxi drivers in in foreign lands are our smartphones automatically identify the phases in our photos and that net progress as you said it naturally leads people to wonder where it will all end will robots get so smart they turn us into pets Tesla founder Elon Musk says AI is Humanity's biggest existential threat and it poses a fundamental risk to the existence of civilization similarly the late renowned physicist Stephen Hawking said it could spell the end of the human race I am very frustrated with this kind of fear-inducing hype in the concomitant overstatement of AI capabilities by by vendors and that's what spurred me to write this book it in my book I explain it in simple terms to mainstream audience yai systems are not going to become intelligent enough to have the ability to exterminate us turn us into pets or even take all our jobs that's that's good to hear a similar point tonight I listen to his attacked I'll talk while ago and I think the basic missed it goes like this the way is it we have and very small amount of intelligence if you see in the machine so the question is of Z how to beat Define intelligence I think we going to get into that but what happens is it's definitely it's rising so something is there that we can finally be headed into 80s probably and I'm curious about that story a lot baby head in the 80s but now we see the sunrise and the question is not will machines take over from Sam Harris's point of you and I think you know masicka makes a similar view it is only a question of when and the in probably won't be in our lifetimes I think that's that's the way we can all agree on this but is it in 200 years is in 2000 years isn't 200,000 years at one point you know we we we can all grief the question is Eid a lot of people work for the AI in or close to that topic if you ask them they would all say this is this is not even in the box right now in this is not something you have to worry about but it's further you move out and zoom out a lot of people say well it might be in my lifetime but it could be and my children's lifetimes and them but it is something we should think about to spit like the AP nuclear nuclear bomb wee-wee nude the basic technology for this a hundred years ago but it took 30 40 years to make it happen and then it would have been good to be prepared to be morally as well as policy-wise prepare for this I think this is where most of the diffuser comes from the sphere is generated in order to generate this the push to create policy you don't is it really like the bomb nuclear bomb or is it better analogy time travel in my view even though we've done a great job with with creating things that work with AI no one has any idea of how to build common sense in the computers and that's what you what you really need to get intelligent computers and all of the all the different AI Technologies we have today they're all dead ends when it comes to this level of intelligent computers so we really have to start from scratch you even though we both great things we can't use any of those great things to build in what some people call human-level intelligence and what some people call artificial general intelligence and to me that that makes this more analogous to time travel you know people have some crazy theories about how time travel might develop some day but do we really would you want to say that time travels around the corner or you'll maybe in our children's lifetimes I think it's about time travel but that wasn't that wasn't real seem like the cold fusion that seems to pop up every 10 years and then it turns out that this is real but that I saw it on the atomic level up maybe let's dive into the first a little deeper I read your book to you the way I want a one and it's a lot about math is laws of physics and you you break it down I think you do a fantastic job there you break it down into I think most readers can understand that. I I've been using a lot of Baha'i tools myself the last 12 months I think this is just was wonderful for me even knowing some of the details I see a lot more I got a much wider Zuma picture and I thought that's fantastic and what what do you realize nothing we both know that is that at the current stage it's kind of a slightly better than a Brute Force guessing mechanism that's how most AI works and that that's not the picture most people have in mind but I think what's the results that this Brute Force guessing generates is sometimes just short of being Magic Yeah Yeah Yeahs NBC would have just said Google and it gives us will be we like my my my children talk to Alexa about a certain topic which is really creepy and then two days later we have an ad about the exact same topic and I'm like what what this is pretty cool the creepy yeah like magic to me and that's what it what I wanted to get it is at what point is you know basic math and basic guessing organisms at what point do they look like magic Sav we have some kind of Technology with Spacek but introduced to someone 5000 years ago would look like magic yeah I think that's I think that's right but you know this will get it from another perspective so in the late 70s when I was getting my working on my PhD at Johns Hopkins in Baltimore Maryland I taught statistics at Towson University I'm and I taught students regression and classification regression being learning from an input table how to predict a numerical output or learning classification being learning from the input table how to protect the classification so for example I remember teaching you know if you had a a big table of historical sale prices and you know you had to calm that had how many rooms in the house and I do with the square footage and so forth you could build a regression algorithm that would predict sales prices for houses that haven't come on the market yet and if you wouldn't be perfect but today systems aren't either and you know what we call it Computing a function and if anybody were to say yes oh those functions have some intelligence in them say they're crazy now those are just stupid functions but actually every AI program today is just one of those stupid functions either a regression function or classification function in the difference between now and in the 1978 is that with bigger computers and better algorithms we can produce we can calculate much more complex functions they can do some pretty impressive things but it's still pretty much you're taking a table of inputs and producing a classification so for example facial recognition is just taking us a table of images import each one labeled with the correct name and then learning to classify an image with that correct name so that if you see another image of one of the people that it had learned about it would be able to correctly predict that person X is that image but that's all I can do only do that one function it can't do another visual function like determining distinguishing a dog from a cat you can't translate language you can't do anything else and if you try to teach him to do something else it forgets how to classify faces and I thought that's really interesting than some of the researcher interviews incorrect labels say it was a monkey with a guitar and it would look at this the first immediate orders that's not a person we don't have to worry about the facial images for for human facial recognition but computers are not able to distinguish. At some point unless you specifically tell them that something you need to train for right yes so when in in that in that example because there has never been a guitar in the in the training dataset or wherever there was a guitar was always associated with a human it is latched onto that guitar and said oh this is a picture of a human holding a guitar not a monkey yeah he's all growns are easy to fool and I think the question is will they ever get too attached to some sense making and I know we've had lines and it was last week it was Deep Mind success in figuring out protein folding which seems to be since then has been mixed messages coming out at this isn't as big a deal as it sounded and wasn't actually such a big step for but definitely it's it's a challenge you've been having for 50 years and also the verb there were a lot of talk that Google duplex AI that they trained was apparently being able to fool the Turing test so do Zara in computer science do you think they've been around forever and it didn't make any progress MLB it seems like if you for looking fitter every other day that is something that has been around for fifty or a hundred years was unable to know perfectly solve it be getting close to solving at that that's that's got to count for something but but what I would say is every single one of these amazing achievement is because somebody's been able to figure out how to turn it into a classification problem a statistical classification problem and if we looked at what these systems are doing as computational statistics as opposed to going to court artificial intelligence nobody would be worried that they're going to take over the world people would just say wow statistics has really come a long way but they wouldn't be worried about the Terminator or are computers turn into send a text at today yeah so we are closed because we're just a couple years behind why do you think that is what do you think we know if you haven't made that 50 years ago was it just a data sets of the data itself has gotten better and more digitized data or is it something that happened because Google needs it I always feel like because Google pixel much money off AI in advertising they just but billions I had it like you know Lefty is BU the public research would be 50 years ago and all the money that they throw it it fourth verse improved over time and we are kind of seeing that in the open source community that the lot of that is actually coming from Google to think that's the driver you know I think I think Google got into it after that happened in and got into it in a big way I think the the algorithms have evolved Jeffrey Henson pursued neural networks Ian McEwan in Yoshua bengio when you're going to miss you over and it it's the neural network algorithm that have enabled us to go from 1978 when I was teaching statistics we were mostly limited to regression functions that were linear now with neural networks we can calculate anytime star classification from now we can calculate classification functions that are you know in a massive number of Dimensions you know hundreds thousands millions of Dimensions you can't even you can't even imagine them but these algorithms have just gotten so powerful that they can learn functions that are very very complex and that's what enables them to do these things now of course the advanced computing power it's also a big enabler and Big Data provides a lot more use cases to tackle but it's mainly been computing power and algorithms that have but I have been an improvement from the early days of linear classification functions let me know your own network sound pretty fancy that's that's one part that still escaped from a little bit of the extra work but the idea is that you kind of breakdown of complex products into less complicated once if that's how you solve these many many dimensions so how does it actually work for underlayment storms I know you describe it in the book but I can follow it to be honest yeah you know what are the interesting things about neural networks is that it's really hard to figure out what's going on inside the network. and the Segway on to another topic we can about a discrimination and AI systems making decisions because especially if the neural network-based you don't know what's going on inside that Network you don't know how they're making decisions and people being affected to them being affected by them these systems to determine whether and get bank loans the facial recognition systems that trying to determine if there if they're terrorists at the airport but we don't know how they're really working inside yes that's I think a big problem all of the eyes that that I've been working with that I broke myself you didn't just two problems long as you don't know how it arrived at a certain look for clues about some black box I think that's that's part of that the game that the other big problem is you always need someone to validate the results there is Sino in build validation efforts that every training comes with but still you need like a somewhat intelligent person typically it's it's inhuman a data scientist who'd and evaluates the results of his or his what I expected and this is why we are actually doing and so do those two things definitely require some common sense that is rare in in the wall of the I am but but I'm wondering is it is that really something because that will only be with humans or isn't there a way that at some point you you will have any idea that can simulate, it sounds like he kind of just have that in the Turing test on a simple level why is common sense and I know it is you have been in the book it has way more Dimension was way more complicated problem but if you FB creep came from a simple statistics function and now have thousands wouldn't be able to be able to in 10 years to do something that has a bit of Common Sense described it as a dimension function or something that you know who we just trying to look for the right AIW have to throw out this problem the problem with that suppose you had all these little little little modules that could do that could do things and now you're trying to build a big Marge a little figure out which sub-module to invoke the promise that that control module would need common sense you're just pushing off the problem into that into that control module the alternative is to write a conventional software program will you say if x then y otherwise if Y and Z and so forth in it and that we know from our history from the 1980s that we can't be successful creating human-level intelligence using rule-based systems yeah it's just too hard but if so when will you be back about a ivb have that idea that AI is going to be at the Superhuman the most intelligent person you know when we know it at the speeding chess player but yes. In this consistently the last 15 years I'll be always always felt like chess player than the most intelligent people on the planet and they were like I looked up to them and I still do then the maybe that's not a good idea but that's that's kind of how I felt today and if we we see children you know if I if I have to three-year-old about that common-sense question that might be a certain things that a three-year-old has but in general will be on a beauty limited leveling I think in between me of anyone who is having a high amount of Common Sense who has no common sense and I always feel today I has now arrived at a level up herb teenager maybe not a sixteen-year-old but maybe it call the old and 10 years ago was maybe at the level of a four-year-old so the deed examples that have been made as a GPT 3 is now as good as most teenagers are in school that has a text of essays that it produces computer coated it can generate death it doesn't go to Vail for teenagers at the store now yeah I don't I don't agree with that it looks like those things one one at a time talk about essays so if you look at it the shg pt3 generates at least half the facts are Braun what does it's really understand what it's saying yes I debit that would also look like the most teenagers you know when I when I created essays in high school I researched them and they were fairly factual women they may not have been very interesting but but I was able to create factual factual essays and I think most teenagers can do that in getting back to children no I'm just saying Deedee you know would most like a 5th grader with two or six afraid of seven straight or what they would do is copy and paste from existing material then maybe change to send text with little sodium not as easily detected and that's how I usually do things look like that that's not unlike GPT 3 which is not an original work but it's not fully original item to user but it always feels like it's something that comes from a sore that already existed because that's how does worcs the letter on a second kind of looks like a translating a eye and is and that's how does works when when I look at my children they think they look at they did they maybe got a certain I understand material that's already there then the recombine it to rehash it and then they put this into the form that is Das a or that support I say let's give a short essay or what whatever their past papers about but they but they get the facts right it means understanding the big picture ramifications that they have some idea what they're writing and it gets huge numbers of fast phone yeah but it's we weave that's kind of server problem we we we have something to possess of the seemingly logical answer and it will take a bunch of people may be right now just a few minutes but in a few years it will take them a few days to figure out how to speak correct answer or not that's actually it's it copied it somewhere else like the verification of the answer I think becomes a bigger problem with a eyes as they say progress yeah although it's in a weaving without a & i we have that we have that problem with fake news people read fake news and they don't bother to verify the facts even though the reason we verify Apple what do you think do you think this whole problem with fake news which is a relatively recent phenomenon Lisa PVP have to submit scale is that driven by AI in the sense that you Ai and wavy beep beep received the news. Out the way that they directly interact with our brain is that a is has made us vulnerable vulnerable or more susceptible to this so that way I can already changing our brain but they are that is running the algorithm for Facebook and Twitter and Google it pushes things up that we don't really care if they're fake or not it just it if it generates an emotional reaction with us and we are more interested in this at least for the last 5 years into getting a factor a lot of people know how to validate getting us is emotional adrenaline shark especially when you look at but those algorithms are there to to generate more engagement so the people stay on the platform and click on the ads so that the the platforms like Facebook make money and so what the hell going to send to do is is is Purdue's show people more and more of things they they might be interested in and get them interested in things on it and unfortunately a lot of times that process pushes people into Fringe groups and you know once once you start getting those Fringe groups You're Nobody checks fax is it just is just spreading rumors and you passed the rumors from one person to another and it goes viral and yeah I think that's a big societal problem although I think it has more to do with the social networks than it does to do with AI it certainly the AI but I think it would happen anyway if it's hard to find a way to find I don't know what the way back is because I feel when when we lived in an on social media environment we had and we had the normal distribution of things that could happen to us what I'm trying to say is we we we do things and we feel okay the car driving can be it can be dangerous but driving a car if nobody ever hit me in the last 10 years so it isn't that bad so we have like an intuitive feeling for pretty normal distribution of events in our life but with the social media and the way I has his digested all this and I think social media started was the same day or reflection of the real world has lost things that you feel like they can never happen to you that kind of outside of your sphere AI has moved and resorted the whole world if you something that looks extremely scary that's why being gay true that it's it's scary and maybe in a good way or bad way of I think the battery is more accessible so we always you feel like reading Twitter we get hit take heart immediately Wikipedia normal distribution of experiences as completely obscured by the eye and nobody is able to verify design engineer set out to do that at Twitter or Facebook but it's kind of running now and it's been I think this is what most people feel this is the danger of AI you set this thing in motion and it becomes like a weapon of mass destruction because even the engine is a build it they may be can shut it down but then when does Office Depot pay bill will come up once today I the motion you have stopped it anymore but what if it's not Twitter it would be ticktockers I'll take their word for Youth Baseball it's what it didn't use any AI you get a lot of the same behavior a I might exacerbate it a little bit to be you like Sadie the early days on Facebook it was more boring and there was more that's a random stuff I call it on the side but it would reflect it real life better cuz 99% of what the experiences of its random that's why they just boring Randomness to worry about me worry about the car crash that he stopped and we like what we need to see it is because then we change our picture of reality nothing good happens right now that the oldest I was so you can say that social media that someone would have used they are for this may be enough now but the way that is changing our perception to be constantly have to adjust our men image of the world all the time because of always a car crashed at 1 in the morning when I open my Twitter does one of the evening and then does one for lunch at these things would only happen to me and you know I don't want to hear that would see the trash out why do you why do you blame that on a I am not Twitter what that blame it on the eye but I feel like I got is Gaba so good at that job. This is Gospel is one of a problem that has already changed the life of billions of people and I don't think Twitter has used technology to something evil you can put it this way supposedly didn't use any AI technology they just use conventional computer software was if then else rules they could still have done 90 95% of what what happens if you still have listened people to follow you still have your still have algorithms that will show you the viral posts you'll still be able to still be able to figure out things that you like in the eye shop is pencil a little bit in and and you know makes all these traditional softer algorithms a little more pointed butt I don't see it as the AI that's causing these problems I think they would be there if if if the social networks didn't use any II whatsoever what's Italy at the margin I agree with you at some point it becomes definition problem because this is not like a self living being has become conscious and I was going to take over the world that's for sure that that's not what it happened how it happened or will happen at any time soon but that is something maybe it's just scale maybe it's the sheer scale of you just need a bunch of computing computers and every kids can have an impact on the war of mine to are in a matter of days if you have the right away I so for a IV use too much afraid of supposed instead of AI we started complaining about statistics you know statistics is taking over the world status Twitter is using statistics to make the inner tube pushes into French groups Facebook is using statistics nobody would get very excited when you say it's it's using AI to do that now people start worrying because of the led the idea of AI is is The Terminator maybe that's the fours that tree issue with people who are close to a i and work for them every day to think it's it's just an extension of the Alico and I think that's true did this you cannot really dispute that but on the other hand we see that this is progress that has two years ago that Tesla said we going to have self-driving cars cars on the ground in Phoenix that are supposed to leave out driving is sometimes a driver in it or like as a safety driver or at least eight Dave made that announcement the most people who sings out driving cars for 50 years problem that never get solved and suddenly it's there I mean it's it's kind of exciting and you couldn't it's easy to extrapolate this into the next 10 years yeah yeah yeah and it is it is it it is interesting because the self-driving car companies have done amazing work I mean I I have a Tesla and I run it in autopilot you know it is probably does 90% of the driving but if I let it to 100% of the driving out of smashed that care about a hundred times over yeah so but it is amazing what you can do with with the self-driving car technology in that that is a great use of great use of AI and we're seeing self-driving cars be practical today in some in some areas I mean today you can go to corporate campus and you can see a self-driving shuttle go from point A to point B with no driver and the reason I do that is that it's traveling at 5 miles an hour the route never changes so it knows everything the other designers know everything that might happen along the Route so there's no huge need for common-sense decision-making and at 5 miles an hour especially if it uses computer vision to make sure it doesn't run over pedestrian is very little chance of a major injury now you go up to scale a little bit and we're starting to see delivery vehicles so in some cities they're allowing self-driving they look like little ice chest that are moving along the the sidewalks these computer vision to avoid bumping into people and they deliver things and again same idea it's a little bit harder because the more things can happen and there's more to rain differences in that they don't have the ability to use common sense so if they get into a bad situation they probably just stopped and I can't imagine that happens quite a bit and then the next step up for the self-driving taxis which are being tested in cities like San Francisco by companies like Zoosk and Aurora and in Phoenix by Google's waymo division in what was happening there is there their they're all being tested in very small areas where they have every little every stop sign every fire hydrant every work area everything completely mapped out and if any of that changes the cars will get stuck or have an accident and they're doing it when was doing it in Phoenix with safety driver's the ones that don't have driver's my understanding is that is a remote driver you open drive the car with a joystick if necessary but I'm very worried and if you talk to people I don't know we went we may have had this conversation cuz you're from the San Francisco area that the self-driving taxis are often blocking traffic because there is so conservative and then people honk at them and they go too slow and was it you who was telling me that it was that somebody that we talked about that I think last time in DD the waymo cars with most of the men will cars be counted as we still have tons of them in the city then they drive like like a grandfather you know literally on his on his last day of driving before he gives up the driver's license it's it's extremely cautious it's an in San Francisco the density of a lots of pedestrian traffic and everyone is pissed people on skateboards and scooters at does a lot of stuff going on usually people watch out for each other but the idea of the algorithm is to be to do Safety 1st and I think this this is big deal how you prioritize speed safety all of those things that go on automatically once you are experienced driver friends in the city of San Francisco and what happens is the is the people who are cautiously into the street because they want to cross but they wouldn't go you can see them suffering was cars with just keep going to stop at a stop at far away and a lot of accidents of being produced because they stopped in the way that even student drivers were extremely cautious they get confused as you say but a little this changes we have a lot of steep hills and sometimes they usually have to have a driver's but it takes them a few seconds or sometimes if they are not about to fall asleep sometimes with all this and so white I see it the day today but I do I do feel it's gotten way better so we see a lot more of those cars and think they still have to save the drivers or drivers just put more effort in but to me it seems like there's less and less if that's annoying stay still around but they did it it's driving much smoother than I've ever seen before in love people called us to search for and what did you just need to map out the whole world which Google is it has been doing the whole time the Google Maps for Resident BBB have those the street view images so if you assume you have a perfectly mapped environment the self-driving cars might not look that bad anymore. The question is will they ever get to 100% of my take a long time probably would love to get to 99.9 in most situations may be noted that City like San Francisco most cities in the world you know what I can see that happening very soon maybe the next 5 to 10 years. Besides a few cities with lots of traffic dos things will just do 99% of the driving if not more but what good is 99% of your still have to have a safety driver that's a good question that's a that's a good question I mean it still helps right on the freeway but it's it's not enough people ever get rid of that the safety driver for the next 20 to 30 years equals like Tesla's innovate they don't have the advantage of of knowing where every at stop sign and fire hydrant is and work area up because I can drive anywhere so that makes a problem that much that much harder and there were a much wider range of things that can happen you know black ice a ball bouncing in the street and with the child following it that people use a common sense to figure out what to do and cars can't do that you can you can write a conventional software program that says if you see a ball bouncing in the street. Because a child might follow but that's very different than what a person does people people don't learn all these rules and driving school they they used when they have counters in encounter situations that use their common sense so if a car can't have common sense it's it's hard to imagine how a consumer vehicle like a Tesla could ever really get to level 345 driving capabilities level 3 meaning you no longer have to keep your hands on the wheel you can watch a movie or read a book one thing that they definitely have on I don't think this is this is being part of the AI threat that was painted once you have any idea that learned it had learned it to a level that is sufficiently good at it every single computer in the world who uses that model me to learning takes a long time and then eats a lot of gpus and pick machines but once you've learned it and condenses into a model it's it's almost like an instant knowledge but maybe the word knowledge is not correct but it's an instant instant axillary machine in the world that one the weather inside access photos. What we do as humans and you feel those with children they go through this phase where they just like children and then they they they they go through a phase or become more of an individual but they all have to go through the same steps of learning which seems incredibly inefficient that maybe that's not true but it seems incredibly inefficient why not just started a much higher level and me that's that's worthy the old driving debate comes in if if enough big companies work on this maybe they work together in the open-source eventually Supra parts of it then it's listening for metal advancement is it goes big jumps so it goes because you're smart mapping out there that's better trained as there's there's the ball bouncing ball has been Incorporated and Sunday every Tesla in the world or the next time I think this is where everyone is so excited about that once you have a model that works and BB have models that work at least turn off for father and simple things it suddenly is it on everyone's car or in everyone's car and you know from there it only gets better. The other driver I think that's how TI argument cuz you know I've seen my Tesla get better and better things because you will get over the air updates once a month just like you're just like you do on an iPhone but it still doesn't have common sense so if if the car is driving by itself and there's a sharp curve on the highway if I let the Tesla continue to drive at that speed it will go right off the road yeah is that is at the training is short data issue from your perspective so how did I learn to slow down on curves had an accident but I can tell you that and can you take that one situation and I'm sure that test will eventually know how to slow down going into curves either because the engineers will write conventional if-then-else software or because still feel train the system sound machine learning algorithm and he'll train it on a lot of a lot of Curves but the two things that one the person only takes requires one example to train the machine you need thousands tens of thousands maybe hundreds of thousands of examples of of those sharp curves and then you have all the other all the other things that happened to you you know in it when you're driving $50 almost anybody will tell you about driving situations that they think is a one-of-a-kind almost everybody has their stories how are you going to train in Europe there are four billion people in the world and each one has their own driving stories how you going to get all those into a computer yummy just ETF the problem is described as real that the significance of data that we proceed from like single or event is not something that I think any AI has right now so DD you can obviously attitude a model but the way to do have a looks like we can immediately feel like we are in danger and then immediately to the level of significance jump so high. We will never do it again as you say but I think isn't that just a data problem with again if if we would know the significance of so high and be sure to learn from one experiences not that machines can do it right you just can't see to see the relevance right now which doesn't mean we couldn't find a relevant to all over them so to speak what's the posy word Suppose there are 10 billion of these unusual use cases out there in the world how do you identify all 10 billion and we have this usually it's like your fear like an emotion comes up with it and it itches it's burned in your memory feces life-altering events where do you think you're going to die for a moment this is I don't know I don't know how it works but I'm pretty sure that works in a similar fashion right beat me immediately before we even consciously know what you mean really know her life isn't danger that's how we assign relevance I think there's no scientist today that has any idea how to do that immediately had that sense off you notice fear this Justice you hyped up on adrenaline and that's before I could even understand if I'd injuries what happened who was at fault what actually happened was the car talking to me was I running into someone I know this is like a life-altering so I just need to be as awake as possible. That was definitely not a reason it come in and everything else but reasoning let's put it this way I think if we could build an order that way when you say it wasn't about reasoning so so what did you do what did you do that wasn't reasoning what kind of dentist my the side of my in my car and I ran off to them in the middle of traffic is the opposite of reasoning I was high I was high on. I would argue it was reasoning that you know maybe it was faulty reasoning but there was a lot of reasoning going on there I mean just just starting with a how you get out of your car you know so you know you you reach with your hand and you know if you apply pressure in in in you you grasp the handle and you apply pressure the door will open then you know you have to push you know all of these things how do you know all these things and then you know how to how to how to run a lot of lot of common sense reason and Common Sense knowledge and reason that you do get to do all of that and then how would I how would you get all that into a computer oh yeah but that's triggered by the sense of attention attention was there before I did any of this but put it this way I mean it it just happened like I don't know half a second less than half a second I was probably I barely have any memory so I'm not really consciously aware what was I thinking or what was going through my mind I just saw a protective mechanism from the limbic brain but if the limbic brain can do it and animals can do it three hundred thousand years ago I think we can do it to that doesn't solve the reasoning problem it just causes immediate problem of armed and dangerous I need to pay attention and that I pushed higher in my in my learning priorities data set then you'd might learn incorrectly and that's the thing you back to the 8th and then problem which best travel computer science are you a fan of Ray Kurzweil do believe in the singularity do you think that makes sense what he says and do you think we have this unlimited amount of computing power and light just 23 years from now you don't like I don't agree with that so if you take a computer from 1980 amazing statistic that today's iPhones have as much computing power as a cray supercomputer in 1985 do it if you take an old-style computer from 1980 1985 and you put a word processing program on it the only thing I can do is word processing if you take out a really powerful computer today that's you know billions or trillions of times more powerful than that 1980 computer and you put a word processing program on it and that's all you put on it the only thing you can do is word processing now if you make a computer that's a trillion times more powerful than the ones today and the only thing you put on that computer is a word processor that's all it's going to be able to do going to be able to do it really fast but that's all it's going to be able to do yeah but will yes you're absolutely correct with we've been we've been putting other stuff in the word processor Arizona rifle Supply. I mean I think the iPhone is less of the America land this it's the server Parts the cloud wear a lot of the islets that they speak Russian language data science and comes up with patterns in especially unsupervised learning you have that in your book sock patterns that humans would not see it right away because of too many dimensions or it is just a week on duty parts that they played us at but for AI becomes visible these patterns and I think this is very bad to get sweat this is this is ready to self driving cars come from they didn't come from something that did runs on the front end it is really something that that happens when these several parts of the old couple of python modules and the most a Lennox I think this this is where the magic happened in the last 10 years and review scale keeps getting this and just thinking linearly because will pass superhumans I just think linearly that could be pretty amazing already and if you think this is like worthwhile says is like a doubles and then he said to eat that's why he calls it the singularity he says it's going to be so amazing it's kind of a hopefully optimistic of you point it's going to be so amazing that it's all the people look Beyond at the tennis all our problems it's it's become verified. But I think it's such a hopeful message and it seems to make sense just in any Court statistics and I agree with your point it if you don't develop the software if you don't give out the right mindset is not worth anything if you just built big machine to kill us War Machine weapons of war then it's not good at all but if we managed to use it for something useful Maya question to this is usually entrepreneurship it with entrepreneurship can only live if you find someone who buys your stuff with you just make things that nobody wants pain and nobody blocks improved then you're not allowed to bring her I mean you tried but it successful entrepreneurs and it often takes a lot of a lot of things to learn eventually make everyone's life that touches your solution you make that around a dead person makeup on her decision to improve the life let's go back for men at the house self-driving cars work because most of the AI self-driving cars is a set of supervised learning algorithms so you've got a a program that can recognize a stop sign that's one supervisor running out of it's been trained is in the car and knows how to use it and there are about fifty of those recognizing pedestrians record their other other supervisor and programs that you know figure out what their trajectory is of this person or this are in a second from now so they're all these little all these little individual programs in there and then mostly connected by conventional if then else programming is very you know I don't know I can't think of it I don't know of anything in a self-driving car that is unsupervised learning where it's going out and and figuring something out it's a very specific classification algorithms connected by procedural code who were the unsupervised learning techniques have had impressive results are in cases like GPT 3 where you give the computer a supervised task of predicting the next word and you get you get some interesting results how does a lane changing our growth of work always thought that sounded Super Wise eventually you know I'm I'm not I'm not sure but I can't imagine how you would do that and unsupervised why I would imagine that we need to be a supervisor learning algorithm or reinforcement learning algorithm that you know each situation is is labeled with the with the correct response and actually I think I see what you're saying so you could look at you look at Lane changes as soup as unsupervised learning in in one respect which is that in every situation you can take the correct label to be what the actual driver did say you're in a Tesla in the system can look at okay did the person change lanes or not and what preceded that but that's that's self supervised learning which is considered the form of unsupervised learning but that's really supervised learning where the labels are provided by the by the environment it woke did this labeling is at the seams right titty titty is the whole idea that we have to run through the cell labeling process so I did the things less less impressive than a new patterns of refined unsupervised learning to be honest I don't know the exact boundaries between the two it seems like they did kind of shift but it keeps your thing because nothing is ever the results kind of shape the how Bella does results are they seem to shape the kind of like labels in my mind at least so when I when you come up with something that that is ways to overcome these hundreds of thousands of that mentions on Netflix have big issues with this and then they came over the way to move that into classes on Sunday become way more easier to handle that's going on and all the dimensions and about driving cars because of so many inputs at any point of time at all of them could be a poor but the ball bouncing in front of the car which has no kids around 2 maybe it's not a big deal to run over the ball yeah yeah I mean a lot of things here but that's kind of what I thought you know I don't I have a practical idea of how it works for me at the pond scene from a python script playing with some dataframes but if you hire application some office Lee and not an expert at all yeah you know that they're at one of the one of the theories about how to get to human-level intelligence and it's one that's pushed by some of the big names in a I like young looking and yahshua Benji oh is that you can give a computer a task like learning to predict the next word in a text with the idea that what's going on under the hood is that the in order to do that task the computer is learning way more than a simple function that it's learning knowledge about the world it's learning how to reason about the world and and so forth but you know to me that's kind of wishful thinking and I think the evidence of that is that it learned so many wrong facts and this is much easier explanation of GPT 3 witches that is just piecing together words and phrases that it's encountered in the in the documents that it's it was trained on that's probably true. I guess we will never know that's the problem with the whole validation right we don't know what's inside how it learned how it came to Circle crop photos on my favorite my favorite question is asking people how do you how did you come to this conclusion what what changed your worldview what changed your view on a specific question the day I is you run against the wall that they even at least the current Eli's they can even if they would have a language they can't tell you why they arrived at the same conclusion I feel that's very disappointing right right it's it's it it's disappointing and it's a it's a big problem for society if you don't know you don't know how to do dishes priorities we don't know how they assign weight of a competing priorities of like for driving everyone says we need to open sores that the dog with him because I had my running over to you at the desk in front of me or I'm going to break and I'm probably going to kill everyone inside the car and for that do I have to count the numbers will pass just inside and pedestrians or do I go by age and if The Pedestrian is really old and I don't I mean that stuff gets infinitely complex because it has a probably more data than the than the algorithms I want to lead to a different Rivera related topic and I keep asking a lot of people on the reasoning so interesting you know it has been distant this debate for some time the paper came out about the decade ago if we live in a simulation it's kind of a trove now how do you feel about that how do you how do you think that makes the original the outside of that paper makes sense and do you agree with it you know this is really no way to tell I mean if we were in The Matrix we wouldn't be able to know that's why it's the thought experiment fugitive do you feel like instinctively from what you've seen we could build a world in say a thousand years from now or two thousand years from now that would be so indistinguishable from a real world on higher-level like we could build a whole universe because eventually it was just to know the size of SD card in the we create the universe expanded could we could be created or that's something that will always be like in the sphere of religious person would be kind of put in these brackets we can't describe it but it's someone with a nightmare but I'm getting too has this whole to hold universe and AI to is it Guided by someone with G+ message towards our future to watch the human future or is it all right then and there was never a simulation and there's nobody steering this whole thing right you know I think there's two questions there one is are we inside a simulation and I don't think we can ever know so I haven't spent a lot of time thinking about that the other one that's more more interesting to me at least is could we ever create a simulation like like the Matrix like that the movie The Matrix in the movie Matrix and it'll work we're just starting to see you know Elon Musk is just starting to interface computers to the brain but we're out we're a long long way from figuring out how to make the brain cereality I imagine it's it's it's possible Sunday I just don't know yeah but do you feel do you feel this universe that we humans have evolved as someone is driving this car someone is is guiding us along the way might that be spiritual figure might that be aliens do you think there's something to it or that's just know it's a good whiskey if we might not be the only ones but we kind of philosophical. Question you know in around 1960 470 Arthur kressler Road of very well-known book called the ghost in the machine in the idea is to bring forward that ancient philosophical debate about whether is a mind that separate from the physical body it is the mind just you know the result of a collection of neurons or is there something else and I think I think Kessler's position was the materialism position which is just a collection of neurons but just as an interesting aside or at least it's interesting to me I really wanted to name my book there's no ghost in the machine okay the take off on Kessler's Kessler's book applied to computers if I thought was I thought was very clever but my publisher wouldn't let me do that because you said that they never find my book if somebody searched for what one would think a lot of people reading a book will be interested in having three we had to bring the year building up to the election it's the whole definition of gravel or drops be in the future if a guy takes so much of it away and buy of the you know that the daily mail from a car is a big Topic in this even if they only get to 99% that puts a lot of people who are the freeways Auto out of their job they might pick up city drivers but this way Euless city drivers needed compared to what the whole driving trucking business is not requiring from drivers when I think a lot of jobs will get away maybe just let us have to be for the eye to stay at the technological progress and now finally it seems we've covid-19 light up thing a lot of progress much quicker I think it's a good thing but where do you think add it to this will that be the new opportunity so where will the job sprout a armor take away what was the first of all I don't really believe AI is going to take a lot of jobs I think conventional computer software has been taking a lot of jobs over the years in over the like 50-60 year history of of computers you've had word processors replace secretaries tax prep software internet travel sites this place travel agent e-commerce is is killing brick and mortar retail all of those are conventional technology and the history of automation is such that technology has always created more jobs and it's replaced always you know and in 1776 farms and Floyd 80% of the people and now we produce more food with only 2 per Cent and Floyd and out agriculture but people do do a lot of other things so you're the real question for me is whether AI is going to change that that historical Trend and I don't really see it I don't really see a i taking a lot of jobs eat the jobs today I was going to take are the ones that can be classified characterizes classification tasks you know and it would visual classification test being especially vulnerable so spawning terrorists and airports reading MRIs sorting Parts in a factory no voice recognition technology is impacting customer service jobs that involve pharmascript because people can just say the words in the computer will recognize the words in and can follow the script itself but I don't I don't see these as really having as big an impact as traditional computer software and traditional computer software for 60 years has created more jobs than it's taken away and I don't see any reason why that wouldn't continue you've definitely with you on that part that there is this this process of moving from a low productivity to a higher productivity job it's it's still intact nothing has changed that the trouble is that we see deconstruction immediately and of the factors affect the first and then the build-up to what are going to be those new jobs it's it's not often. The clear and you know it is over the anxiety that you might have been a winner in the old system but you're not a winner in the new system so I got that anxiety is always felt as a negative emotion so I think that's what rattles people especially now that we receive such a high adoption rate of new technology I think you can purchase with pastic that's what entrepreneurs have been dreaming of for the last 10 15 years because a lot of things that I personally my startups put into this world in in the 90s late 90s and early 2000 they've earned Rudy adopted widely and that's because of this product was crap but even as a technology it wasn't pretty adopted like video conferencing was technically possible in the early 2017 was already picked up and now it's happening some every once and never once daily agenda and I think this is fantastic that we received this this pick it up Shania and to be honest and I'm not so sure how AI as a definition of a field goal will change that and how how many jobs will you be you say oh this is computer software and this is like Advanced Computer Software and changeable at the margins so I agree with you there. Maybe AI is it as a possible to extremely over-hype but as it is often happening in follow the same path of the technological destruction one thing that I keep thinking is due is that a i as it gets better decision-making you can say there's a lot of stuff but I think AI especially gives us a way to see patterns and to kind of be ahead of of a customer's potential customers preferences before the customer even realizes he or she has these preferences what I mean by that the example is I always felt like we if a guy takes take so many basic decisions I'm out of the equation because it becomes cheap and we cheaper than a human to look at certain data it's better to have a I do twos and eventually eyes good enough before a lot of fields of data humans can focus on his next level of precision pain better anticipate customer demands and always had like my example is someone comes into a bar and there's never been in that bar and the bartender day I looked at them and say this is your favorite drink always felt that that's how I feel Eli's going to change that game yeah yeah it it it that that is how it will change it but is Richard is it just a is that just contributing to the same trajectory that computer software has been on for 60 years of that I fully agree it's an extension of that but I think it might give us the given what covid did it gives us a lot of momentum and that momentum be feel the destruction for adults that's really Factor people's minds together with social media it is you know to me the big the big issue with jobs is that losing a job is like one of the worst things that can happen to a person and it's a society let me know if you lose if you lose 6% of the jobs every year but you create 8% new ones Economist would say that's great but it's awful for the people who lost their jobs and it's a society I think we need to take more responsibility for your retraining those 6% so they can take some of those jobs in the new 8% and you know we've never really done that governments aren't aren't you taking the lead in that and it's kind of its kind of been you know people lose their jobs are the losers and I think that's unfortunate I would I would now we're getting into politics but I would be very much in favor of some kind of tax on technology companies that went to support people who lose their jobs because of Technology I could find that's a good point tonight one thing that I read this book ended the main thesis was of unfortunately forgot the title of it is that while society as a whole has gotten more productive and it's gotten better at extending this is Heidi's of the individual's life span for the individual deed the freedom and deed the life itself might not have got much better and DDD idea comes from think about the life of a typical hunter-gatherer that's how far back does book went and then how the live immediately at least for the beginning of that phase in a more Agricultural Society changed it wasn't the same kind of freedom for the individual life decreased maybe that's true in Edward PDF the quality of life degrees and it's different for the for the Next Generation X Generation starts at a different point of o face blind and you add more Generations it it becomes better but for everyone in between it for that individual it seems like the quality of life is decreasing similar Junction right now we feel that like we had this discreetly called the boomer generation who who came out of the second world war and build wealth 70s 80s 90s great productivity growth at least initially in the 60s and that technology took off and just accelerated that but but it has hadn't happened in the last 40 50 years of age specific automatable is that the SEC Nation came about and we don't have as many opportunities as we had the VC Mimi had this special day for young people one of them but do you know what people suggest as well it's because software and AI is a subfield of this is getting so good that you need way more experienced to be better than the machines WIC so little opportunity for young people because they are not as good yet at making good decisions as you noted the CEOs have gotten old and older they're still there now there I think the early seventies as an as an average age that something you've never seen before presidents are getting older and older Soviet that juncture that is like a promising volt behind that maybe that's happening in 20 or 30 years but for the generation that's kind of in between the Millennials and generation just before that it might not be what if I say the jobs are being created they might not be as good in those jobs you know the 6% who got out of their jobs the 8% to come in there might be different people I would say two things first just a comment about quality of life nobody's going to convince me that my life was better before I had a remote control and I had to walk up and manually turn the knob on my TV. and you know this got to it to it till I know what generation person Gen X Y Z it's going to sound like an old Boomer talking but we always had the idea of paying your dues so before you get into a profession you really had to go in start at the ground level work your butt off and learn something and you know what would some people are saying about the the newer Generations is that they're not willing to do that both of the opportunities aren't there they're just not wanting to pay their dues yeah that's that's that's probably true but I think what the problem is is that a d Millennials have seen that by the time this industry or that company that they invested a decade and by the time it Dave Mitchell maturing into a more senior position and it pays off these companies are not around anymore yeah that's not a false assessment I think that that's spot-on to isn't does exceptions to that rule but in general I think that's a spot on observation is that obviously Google so wrong but it does a lot of technological change. Just look at the apartment of biggest companies in the US 20 years ago and now that is a lot of companies left and I think this is a worldwide phenomenon so does this long-term investment that the Boomer model where you kind of willingly torture yourself for a while because you go in and learn everything from the ground up and then but the company secures you are available Yang once you mature within the organization with you you do lateral changes I think this model is gone and the demotic we are now going towards that you have one grand idea that you leave one thought in your whole life and you wait 55 years for it when you make one app make one call to your broker whatever that decision is and then that's it then you never have to work again but you have to wait 55 years old and I think that's what it feels painful yeah that's an interesting perspective but I agree the rules of change I mean it's there's no more safety and in working for a big company anymore starting out of the bottom and and working your way up but I think the the concept still applies you get in and you learn something you do you learn how Insurance works or learn how we can do whatever industry here in you learn how it works and that makes him more valuable for the next company so yeah you may not be paying your dues and working your way up in one company but you can you can do it going from from company to company or or you know even being you know what individual contributor so I think you know my sense is that the opportunities are still there I think we have a big problem with wealth inequality that you know we're not we're not we're not addressing and that that's impacting the Millennials and I don't blame them for being upset about that but you know talk to look at look at the world and say there's no way to get ahead I just I think that's just a self-defeating attitude yeah I mean I'm I'm I'm in in gen Gen X so I'm just at the at the border but I agree with you in the assessment that wealth inequality where we left me at the turning point of the solution of busy for me and that's a big reason why it was podcast is figuring out where are these opportunities and why haven't we created more opportunities the last 20 years especially I think the 90s was too late 90s was the exception to the rule in the last 50 years why didn't that happen in the country that's so open to opportunity like I would like the US and how can you fix this going forward and what my answer is obviously entrepreneurship as as a way to change the world German by technology because this is where does productivity boost comes from that makes it all better off and have a better life for all children then I don't know if you have a specific ideas where you feel like even if a eyes overvalued as a high-powered where do you feel a I've ever really make an impact of what are those opportunities for entrepreneurial satanic 5 to 15 years were something is bubbling up but it's it's not yet it hype machine W Weed Everyday yeah so you don't know what I said when I tell you why I'm talking about Ai and not traditional computer software so again I think we're AI is making a difference and where it can make a difference is in taking over those tasks that can be classified that they can be characterized his classification tasks so that that's we have to look but to apply that technology is an interesting problem I think some companies say okay I'm going to go higher the best machine learning expert I can find and they're going to bring an AI into the company and transform the company well that's not an approach that's going to work you can bring in the smartest machine-learning person in the world and they don't have no idea how your business works so if you want to bring a into your company you've got it you could have somehow educate your product managers and your business decision makers what's possible with a i and everybody's got to put their heads together ideally with a machine learning expert or the data science team to figure out how to play a i to your company one thing that they crossed my mind a while ago is something like a Boston Consulting but just for EI cuz I think that's probably ton of people who already do it is so I'm just not a new idea but a way to maybe not charge $2,000 a day but maybe charged $1,000 a day but finding that unique dataset I think this is what it's all about finding at any theater Sutton trying a bunch of different learning models and then Valley dating them this this is relatively easy to do if the data can be found within a company but you can have immediate effects when you can see results in terms of conclusions that maybe people know intuitively but I've never really taken a look at it like they've suddenly realize oh only dentist from that area actually subscribe to our product or because this is this is only relevant there we have no competition. I think there's a lot of inside but it takes a while to find out data so I think this is the problem generally with AI is is finding that they decide an iguana large is fine enough and big enough to attend Valley dating whatever the learning outcome is that's a big challenge unless you say you need that industry knowledge and right start up companies going to be successful just bring in a consultant or data science and machine learning expert and and turning them loose I think they have to commit to learning what's possible with a i and you know it lie I think the business has to say to that data scientist but we got all the state about our customers can't you figure something out about the customers that will help us sell more to them and you know then you put the new released pointing the data scientist in the right direction in the business team in the IT team don't know where that they it is they know what's good about it and what's bad about it state of scientist is not going to be able to figure any of that out it's definitely compartmentalize I hit my microphone out on the lot last podcast we talked about how he I can solve so many things in the medical field and health care but how difficult it is to get any data invite enough paper said and even most studies that are running the hospital to have to maybe $10 and maybe fifteen thousand different screening for cancer screening and it's already used made us and it's very expensive to get to replicate into another studies where you're going to run an IMEI on not. That's really sad because you could have such an impact on people's life expectancy if you could apply decal but it's just not there the dataset in a way that you can easily access it and that hopefully is going to change Monday it would be it would be so beneficial to let me know I just get frustrated when I go to my eye doctor whose wife and saying for for 35 years and you know when I say to him what is that what did that test show last year in two years ago in three years ago he starts paging through his hand scribble notes and what did I say it was two years ago I mean it just you know you don't want to answer the question yeah I think that means the healthcare field has completely missed out on I mean software automation the first place that slowly now making its way into the field but it's definitely not eating Healthcare is it that all the other fields any I just hasn't arrived at yet people don't understand the value of that day. They think it's just random and it's just there for just one particular moment and then it just feet away didn't understand that each data set a date of each piece of data together has a meeting for all Humanity so to speak because if you have enough day that you could see the patterns why some people have to get certain diseases and others don't exactly exactly it is it's it's a real shame that we can't we just can't seem to put that together so maybe an opportunity. Yeah it's a the big opportunity that was Obama that was one of his Cornerstone initiatives to try to get all that data together to to build the what did he call it that the home if I forget what he called it but they weren't very successful with that yeah it's a very difficult feel to be at the Disco many legal issues so many people did somebody Gatekeepers in that field got to find they call it the beach at that easy place where you can build build your company and then go from there once you have enough like oppression but nobody has been successful yet but it's it's definitely happening 23andMe is one of those companies who hunted beach at probably have a lot of money from Google but date Dave the hover over this DNA tested and I think ancestry and doesn't like every country has a few companies that do this exact same kind of test but I don't think they share any data so the data cannot decide to still Limited well that definitely found some opportunities in this house that I really want to thank you for coming on that was really interesting thanks for sharing your thoughts with me it's been a great conversation towards and I really enjoyed it