Select Page

great to be here seem to be echoing I used to hang out not too far from here on Canal Street where I would buy something about this big was a telephone really for $50 it could actually switch a bit of information and I was excited in 1968 that I could buy a transistor for one dollar that did the same thing today you can get several billion of them for a dollar and I’ll talk a little bit more about that exponential trend but how many of you are familiar with the TV game Jeopardy okay so tell me if you can get this jeopardy query the current President of the United States who raise your hand if you think you could respond to that okay so what who is Barack Obama who is Barack Obama so you seem to be pretty good at this game how about this one a long tiresome speech delivered by a frothy pie topping the answer is what is a meringue harangue how about this one a garments worn by a child perhaps aboard an operatic ship huh you got it what is a pinafore me anyway Watson the IBM computer got these correct and the two best human players in the world got them wrong Watson also made some stupid mistakes but Watson got a higher score than the best to you and players put together and this is not a narrow task it involves understanding pretty subtle forms of language metaphors puns similes jokes and involves having some command of all of human knowledge and the knowledge that Watson has without hand programmed by the scientists in some language like Lisp Watson actually got that by reading Wikipedia in several other encyclopedias 200 million pages of natural language and it can kind of go through all of that in three seconds and respond to these queries I made a prediction in the early 80s that by 98 a computer would take the world chess championship and I also predicted that when that happened we would immediately dismiss chess as being of any importance and both of those things happen then in 97 deep blue defeated Kasparov and people said that’s not a big deal of course computer can play a good game of chess at the logic game computers of logic machines that’s to be expected but what they will never do is understand the vagaries subtleties ambiguities of human language that is the unique sort of province of human intelligence Alan Turing based his test as to how you can tell if a computer’s operating at human levels based on human language so this is a pretty significant milestone that a computer can actually master this language based game which is filled with these kind of very subtle forms of language so AI which is maybe another word for one Enriquez big data really getting intelligence out of all the information is out there in the world is getting some real traction it’s an interesting observation on human behavior that when I make these predictions I predicted systems like Siri and Google now the Google self-driving cars in my 1999 book that these would emerge around 2009/2010 and people thought that was nuts now that it happens people just shrug their shoulders figure they’ve always been around I mean that’s this is how technology evolves when something first is introduced like Siri people say it’s not a big deal doesn’t really work that well you know it’s make it’s making all these mistakes a few years later it works really well and then people says that’s not a big deal that’s been out for years but we’re making exponential gains in both hardware and software I’d like to touch on that in the interest of time I won’t say much about exponential growth to the relief of the you’ve heard me speak before but it’s hard to avoid the concept because our intuition about the future is not exponential it’s linear when you walk through the fields a thousand years ago so K that animals going that way I’m going this way in the path we’re going to meet up at that rock I think I’ll go a different way I turned out to be good for survival these predictors of the future got hardwired in our brains that’s why we have a brain is to predict the future to anticipate the consequences of our actions in inaction but those int built-in predictors are linear we assume that animal is going to keep going at the same speed that was a good assumption but the progression of information technology is exponential and I’ll say just a word about what that means but here’s what I wanted to cover today any questions on any of this ok since they’re no questions I’ll keep going well this is worth looking at what is the difference between linear and exponential progressions linear progressions go 1 2 3 4 5 that’s our intuition at step 30 you’re at 30 exponential progressions go to 4 8 16 doesn’t sound that different but it’s step 30 you’re at a billion and it’s not an idle speculation about the future I mean this is several billion times more powerful than the computer I use as a student per dollar it’s a hundred thousand times smaller we’ll do both of those things again in next 25 years which gives you some idea what will be feasible so this was the first graph like that that I encountered that I came up with 1981 looking at a way to tie my inventions because I realized the world would be a different place by the time I finished a project this is a logarithmic scale as you go up the scale we’re multiplying by powers of 10 so every level on this is a hundred thousand times level than greater than the level below it so this is computers cuts a measure to the power of computers calculations per second for constant dollar since the 1890 census and there’s been trillions increase several billion fold just since I was a student but what’s interesting is look at how smooth and predictable a trajectory that is such as Moore’s law Moore’s laws the part on the right they were shrinking vacuum tubes in 1950s using relays in the 1940s you don’t see World War 1 World War 2 the Great Depression of the Cold War any of the things that happened on this graph it’s just a very inexorable progression people say well it must have slowed down during the recent recession no it had no effect on it any more than the Great Depression did and this is what enables these changes and it turns out that software is progressing in the same way that hardware is progressing so I don’t dwell on these examples of electronics but this is the cost of a transistor one per dollar in 1968 several billion per dollar today they’re actually better because they’re smaller so therefore they’re faster the cost of a transistor cycles come down by half every year that’s a 50 percent deflation rate some of that goes into better performance some of it goes into lower prices that’s why you can buy an iPhone that’s twice as good as the one two years ago for the same cost and where this is headed is for smaller and smaller machines it’s also affecting biology the human genome project was exponential so an example of linear thinking is halfway through that 15 year project one percent of the genome it had been collected so mainstream scientists said declared the project of failure okay one percent seven years can take seven hundred years just like we said my reaction was oh we’ve done one percent we’re almost done because it’s an exponential projection a trajectory one percent is only seven doublings away from 100% and indeed that’s what happened it continued to double every year was finished seven years later and we are now actually reprogramming biology it’s a whole different paradigm for health and medicine and if we had wharton may be doing the discussion we’ll have more to say about that but that’s going to revolutionize health and medicine it’s also going to revolutionize physical things if I want to send you a movie a music album a book just two years ago I would send you a Federal Express package today I can sent you an email attachment and you can turn it into those products but I can also send you an email attachment with a violin or a guitar and you could print it out these are actual guitars and violins that were printed out on the three dimensional printer and these are progressing exponentially the the resolutions getting finer at a rate of 100 and 3d volume per decade the costs are coming down rapidly this is kind of the quiet right before the storm but we’re going to begin to do personal manufacturing with very inexpensive 3d printers very soon with your 3d printer today you can print out 70% of the parts you need to print out another three-dimensional printer and that that will be a hundred percent within five years so I’ve been thinking about thinking for 50 years actually wrote a paper fifty years ago when I was 14 about the nature of human intelligence which I said was based on recognizing patterns that’s what we do really well we’re not very good at logical thinking even back then 50 years ago computers were better at logical thinking than we were and I developed a program that could analyze patterns and musical melodies and then write original music so if I’ve had in Chopin it would actually compose original music they would sound like it was a student of Chopin and I got to play that on a national TV program I’ve got a secret which I know is before all of your time but that was a show back when there were just three networks and I still have the same view in this book now that the human brain particularly the neocortex which is 80% of the brain it’s where we do our thinking is basically a big pattern recognizers specifically 300 million them and I described how this works in the brain but our ability to actually see inside a brain now has also been progressing exponentially these are charts on different types of brain scanning these are logarithmic scales basically we’re doubling the precision the spatial resolution of brain-scanning every year we’re doubling them under data we’re getting were doubling the size and scale of brain simulations every aspect of this project to reverse-engineer the brain is scaling up in an exponential manner and there’s three reasons to reverse engineer which means to understand the brain I mean one is to do a better job of fixing it things like the Parkinson’s disease neural implant which is put inside the body connected into the brain is treating the brain as a network and not just a chemical soup which is what drugs like SSRI drugs do it’ll provide us better insight into ourselves which is the overarching goal of the Arts and Sciences since we’ve had art in science and it will provide us biologically inspired methods to create intelligent machines and this is actually already happening a lot of the methods used in Watson in fact are very similar to the mathematics of what the human brain does which is to build up ideas and concepts patterns if you will in a hierarchical fashion so let me quickly describe how this works and we have 300 million little pattern recognizers so I have a number that recognize the crossbar in a capital A and that’s really all it cares about pretty girl can walk by a beautiful song can play it doesn’t care but when it sees a crossbar in a capital A gets very excited and fires Wow crossbar and that goes up to a higher level another pattern recognizer that’s getting that input as well as other primitive feature detectors and goes on capital A and it fires with a high probability that goes to a higher level that might go of the printed word Apple in another part of the visual cortex there might be a recognizer because an actual Apple in the auditory cortex it might be one that fires it says oh somebody just said Apple go up another 20 levels and it’s getting it’s now sitting at a very high point of a hierarchy and that below it is getting input from the visual system the auditory system the olfactory system and smells a certain perfume and sees a certain fabric it hears a certain voice and goes uh-huh my friend just entered the room go up another 20 levels and you’ve got a pattern recognizer that might say oh she’s pretty that was ironic that’s funny good timing something fell on the floor so you probably think that those high-level pattern recognizers Beauty humour are much more complicated than the ones it’s just recognize the edge of an object or cross one a capital A but they’re actually the same except that the high-level ones are sitting in the different position in that in that grand hierarchy I talked about in the book this girl have had brain surgery and she was conscious you can be conscious during brain surgery because there’s no pain receptors in the brain and whenever the surgeons triggered a particular spot in her neocortex she would laugh and they thought maybe they’re triggering some laugh reflex but they quickly discovered no they’re actually triggering the perception of humor she just found everything hilarious whenever they triggered this spot you guys are so funny just standing there was a typical remark and these guys were not funny so now she obviously has more than one spot that recognizes humor we have tremendous redundancy in general I’ve got lots of pattern recognizers that recognize a crossbar and a capital A but they’re all organized in this grand hierarchy so where does this hiring key come from we’re not born with it that the subtitle of the book is a secret of human thought revealed and that is the secret than the cortex recognizes these patterns that it actually sees in its own experience and then why is it self up in this hierarchy so I offer one year old grandson now and he’s already laid down several layers of this hierarchy we can actually learn one conceptual level at a time that’s why it takes a long time to get up to high-level concepts like irony and many civilizations actually never got to understanding irony which is unfortunate but so we create this our brain creates our thoughts our thoughts create our brain we can actually see this on brain scans and all of these recognizers are the same some of the best evidence for that came out just as I was sending the book out for example what happens to the visual cortex which is this region that processes visual information actually be one particular region that’s the first region handles the lowest level patterns in visual images like the crossbar in a capital A or the edge of an object what happens to that in the congenitally blind person who’s not getting any visual information it actually gets taken over by the frontal cortex which deals with high-level language concepts like beauty and irony to help it with those high-level concepts showing that the these low-level pattern recognizers are actually the same thing they’re capable of handling high or low level features as the case may be it just depends on where they actually wire themselves to be in the in this hierarchy that’s why you can learn a new skill that may be wiped out in a stroke or a brain accident you actually learn it with another region of the neocortex so to relate this to one’s point about immortality I take a somewhat different tack on that we are reprogramming our biochemistry and we’re using our intelligence to do that we’re already smarter than we used to be because we are merged with our technology I mean this device on my belt makes me smarter I felt like a part of my brain went on strike during that one day so quit strike and ultimately they will go inside our bodies and brains because shrinking in size is another exponential progression but I think that’s a distinction without a difference even if they’re outside our bodies they are extensions of who we are but as we actually learn how the neocortex works how the intelligence really functions we can use those as biologically inspired to create machines that will match human intelligence I’ve been very consistent at predicting 2029 as a point where computers will match human intelligence in the ways of humans are superior today and then combine that with a tremendous scale of computers remember Watson actually does not do as good a job as you would reading one PA page of Wikipedia but it read 200 million pages something you and I can’t do and has total recall on all that information so that’s going to be a very powerful combination but it’s not an alien invasion of intelligent machines to compete with us and displace us we create these tools to extend our reach ever since we couldn’t reach fruit in a higher branch who created a tool to extend our physical reach we already extend our mental reach we will make ourselves smarter by combining with this technology that’s the only way we could keep up actually with the exponential growth of knowledge itself and one of the things that we’re going to be able to do is improve our own health we’re already doing that and we will also be able to back up our mind file which is actually a frightening thing that we can’t do that today I you know I’m upset when somebody tells me they’re not backing up their notebook computer their phone friend of mine actually just recently lost a novel because or her notebook was stolen off a train but generally speaking non-biological intelligence is backed up but we do have very important information it’s not a metaphor in our brains that which is basically this whole hierarchy I told you about and the patterns in each of those recognizes that’s actual data we ultimately will be able to capture it back it up restore it if we need to actually expand our thinking in the cloud that’s where we’re headed you know consider that if you do something interesting with this like a search translate a language ask Siri or Google now question it doesn’t take place in this box goes out to the cloud ultimately everything is going to take place in the cloud this is a gateway to the cloud we will put gateways to the cloud in our brains so 300 million pattern recognizers is that a big number or little it was big compared to other mammals only mammals have a neocortex it was the enabling factor that permitted the evolution of language and art and science and invention so it was big from that perspective but it’s also very limiting we might want to have a billion or ten billion or 100 billion pattern recognizers for three seconds just like you might use a million computers for three seconds when you’re doing a complex search we’ll have that opportunity by putting gateways to the cloud inside our brains it’s a 2030 scenario but that’s where we’re headed is that a form of immortality maybe that’s you know the combination of tattoos and big data I think was a very good metaphor for what I’m talking about thank you very much you