welcome back on the rise of ai future space stage it's always a pleasure to talk to and to listen to our next speaker it really is his beard is as impressive as his foresight and pachyderm did good in making him their chief technology enthusia evangelist he is also a sci-fi enthusiast a great writer an engineer a futurist an incredible and incredibly successful blogger who held the number one writer's spot on medium more than 25 times you can ask him he'll tell you he also always changes what he's doing and the messages he is presenting so i'm very curious what he will tell us now he's also a mensch and i like that very much lucky for us he's a rise of ai regular please welcome daniel jeffries yes you have to clap louder we're very few now i'm used to like uproarious applause but uh now we're doing the hybrid thing so there's too much darkness in the world today whether that's economies that are crashing or surging civil unrest or a pandemic that's spun
out of control so usually when i talk about the past and the future i try to give a balanced view of where we've come and where we're going but i'm tired of the darkness and so today i'm only going to talk about the good things that artificial intelligence is going to do in the next decade in the past i've talked about artificial intelligence in 550 and 500 years but today i'm going to laser in on only the next decade artificial intelligence is going to be our interface to the world our most intimate friend it's going to remember to order presents for our best friend when we forget their birthday organize our schedule it's going to change how we do the arts how we heal the sick artificial intelligence is one of these technologies that changes everything and it changes everything because the history of mankind is the history of intelligence all of our innovations come from our mind if you think about humans we're a pretty young species we've been around for 1.8 million years
and for most of that time we were running around in the forest and 99 of our jobs was to find food and for most of human history that's been the case 95 of the people for 90 of the time were connected to the land and getting their most basic food needs met but then something started to change we had the agricultural revolution twelve thousand years ago in the scientific revolution 400 years ago industrial revolution 200 years ago in the information age 50 years now we're moving into the age of intelligence the a of ai the age of ai and if you think about all of those movements every single one of those was an intelligence explosion if you think about the agricultural revolution suddenly people were farming and raising animals so they didn't have to run around and pick up their own food in the forest but there's more than one agricultural revolution and in fact the british agricultural revolution the second agricultural revolution was an intelligence explosion in other words we used to have
half of the land laying fallow half of the land being empty and so in roman times you'd have 50 of the land growing crops and 50 empty you'd do the harvest and then the year after that you'd rotate it because the land was exhausted in the british revolution they figured out how to grow different kinds of crops and grow from 100 of the land which was a huge population explosion and then people began to specialize and specialize they did now only 26 of the people in the world make their living from agriculture and the rest of us doctors lawyers inventing rocket ships and building skyscrapers so things have dramatically changed ai is going to accelerate all of these different industries because intelligence is the bedrock of every one of these things now how are we going to get there the first way that we're going to get there is we're going to need a linux for artificial intelligence and the reason is because we need a platform to to build all these things on top
of artificial intelligence today is very much the province of big tech and r d departments and if you think about somebody like google they go and they hire these machine learning folks out of the university and they say you know roll these ideas into the cloud operating system that all these amazing developers have built for you and the developers say no problem easy to do but that's not the way it works for the rest of the industry we're going to need a canonical stack a software stack that allows people to build all these technologies and infrastructure stack and it's starting to develop now it's one of the things i'm working on with my artificial intelligence infrastructure alliance with 15 different companies and it's still in the early stages where all these different companies are competing and trying to develop the technologies tomorrow we won't have one single platform like we do with linux was a sea change in how we built software when i first started working in linux i had to go in and
talk to different companies and i had to convince them that it wasn't communism because everything was developed in a proprietary way but now linux powers everything the public cloud embedded devices even microsoft which was the bedrock of proprietary software development now you know builds video game systems and has a public cloud and most of their machine learning efforts run on linux and so we're going to need the same thing but it won't be a single stack it'll be a series of different pieces like kubernetes and docker or lamp or mean stack for web development and what that does when you have that bedrock stack is it's going to start to rip it out of the hands of big technology that's going to move to the fortune 500 companies that's going to cascade down to smaller and smaller companies and individuals and they're going to be able to move up the stack to solve much more interesting problems if you think about the way we used to develop software you needed an army of people to
develop an ugly interface for your corporate database behind the firewall and it would get released every six months to a year by the time whatsapp was developed they only needed 35 engineers to reach 400 million people and that's because they were leveraging all this other technology that existed the way we do artificial intelligence now where we're hand labeling data all this is going to look tremendously primitive massaging the data transforming it running tests over and over again manually checking them 95 of these things are going to be automated and the data scientists are going to move up to solve much more interesting problems i'm going to focus on two areas that i think are the most exciting and the first is the arts artificial intelligence is going to radically change how we make movies and music and television if you think about if you think about the way that we do music now for instance um it's it's a person gets them on stage and starts singing just like i'm up on stage and start
talking now but we're already starting to see progress of where we can recreate people's voices open ai published their jukebox were they able to recreate frank sinatra and you know elle fitzgerald elvis presley there was just a if and and more of that is going to happen it's imperfect right now but more and more we're going to be able to recreate these voices from the past or we're going to be able to break apart tracks there was some research recently where they were able to strip out the drums and the voice and different parts and reverse engineer a song and then recreate the individual pieces we'll be able to layer and break those and bend them all together which is going to be very exciting we've already started to see the convergence in in movies of video games and movie technology so over the course of video over the course of my entire life video games have been getting better and better in terms of graphics i've had all the different consoles over the course
of my life although i can't get my hands on a playstation 5 if sony if you're listening i can send you my address after the show but at some point in time we get to a point where we have reached the end of video game technology we're already getting closer and closer to real-time graphics and when you can create real-time graphics and perfect newtonian physics why do you need any new game engines and that's going to begin to change the movie industry dramatically we're going to be able to do real-time actors there's a there's a fantastic book called uh remake by connie willis from the 1980s where it's about a young actress trying to make it in hollywood uh it's a classic coming of age story trying to make it on the silver screen but the big difference now is that there were no actors it was all digital actors the movie studios were simply using artificial intelligence and digitally creating humphrey bogart and marilyn monroe and putting them into stories and so she was
trying to become an actress at a time when there were not now the death of acting is probably overblown we're still going to want to see human performers we're still going to want to see people who can create a breakthrough performance if you think of somebody like heath ledger doing the doing the joker this is something that an artificial intelligence is going to be able to learn nobody was able to think of doing the joker in a different way before he's ledger and now nobody can think of any other way of doing it but ai will be able to mimic expressions it'll be able to mimic movements so you'll need less and less motion capture and animators and more and more of your secondary actors will be ai creations where you tell the machine to act out in a specific way and the animator will tweak it at different points and this is going to going to take movies from something that cost 400 million dollars down to something like 20 million dollars and you'll
be able to create a huge blockbuster special effects with all kinds of digital actors and layer in real actors as well there was also a thing recently where we got a an old record player from my partner and i uh she inherited from her grandfather she's in her 30s so i had to show her how to use it because she never had one as a kid like me but she quickly became very excited about the warm sound of the record player and we matched a huge collection and one of the records she brought home was ella fitzgerald the lost berlin tapes and it was a concert from the 1960s and this concert was recorded by her people and then it was just left in a box and forgotten for 58 years and when we rediscovered it it's this beautiful it's like discovering this brand new lost treasure of ella fitzgerald but what if we didn't have to re you know find a lost set of tapes what if we could recreate ella fitzgerald's voice the
photoshop of music what if we could take the combination of her joyous voice and layer it with someone like amy winehouse and her sadness or edith pioff combine the best of those worlds and have digital singers what if we had emotional meta tags that we could put in addition to the notes that said make this you know sorrowful in this part of the song or soaring and wonderful in this part of the song over here and then it becomes a dance again it doesn't completely put singers out of the business but it starts the singers start to work with things in a brand new way and they start to tease things out in this performance and change it and tweak it over the course of time is ai does the singing for them and of course there's going to be entirely digital creations entirely digital singers but they'll also be real singers too because people love connecting with the real human being if you're working in data science right now you're probably working on getting
people to click you know ads three percent more of the time or tracking attrition but spend time on your passions look into the arts if you love the arts look at the algorithms that they're working on there the magenta project i did a project recently where we made beautiful ambient music with the work that they're doing where it's that using transformers to study the long-term structure of songs look at the algorithms that are being created learn from them make them better work on your passion at night because in ten years or five years the big media companies are going to come calling and knocking on your door to build that photoshop of music and suddenly those skills that you didn't think were very important or sell are now very very important and in high demand now the second area that's going to be very exciting is healthcare it's going to speed up how we create vaccines detect diseases study drug interactions already we're seeing biggest progress in disease detection in 2017 google had an algorithm
that could detect skin cancer at 76 percent 2018 it was 86 percent by 2019 the highest rated algorithms were at 96 percent there was a radiologist writing in radiology today it said there's a new algorithm that he was worried about that marks the end of the specialty of radiology he was referring to lena which is the lymph node assistant from another google research team that had a 99 accuracy detecting breast cancer including on slides that professional radiologists got wrong or only right 38 of the time so certain very challenging cases and he said this is going to be very good for medicine but not great for the role of radiology the specialty of radiology but that's not a bad thing doctors aren't going to necessarily disappear they're going to be stopped working on the nitty-gritty of things we're going to be able to trust the machines to diagnose diseases recommend treatments and doctors are going to spend more time with their patients that's a good thing with the average person gets to spend 17 minutes
with their doctor sounds like a lot of time until you realize that somebody has on average six things they want to talk about so spending two or three minutes talking about each of these things it's not very personalized care if a doctor has you know an mri that's coming out and and it's already been tagged by the ai and they just have to double check it that becomes a lot better in terms of when people are getting this care it's also speeding up how we track diseases so we recently did stuff with the center for new data aka the covet alliance and they were tracking super spreader events using ml and gps signals across the united states they tracked a gigantic bike rally that literally spread corona all over the united states and the midwest but they're also going to speed up drug discovery this is very exciting if you look at the breakthrough of pfizer and bioentech a german company artificial intelligence is at the heart of what they're doing there are but they're
actually not the only one getting to this 90 efficacy we saw that they came out with a trial 90 efficacy they're actually not the only one to do this there are 34 drugs right now in human trials another 145 plus in labs and animal testing now that's astonishing considering that nine or ten months ago we had no idea what the chronovirus was to put that in context the fastest that we have ever had a vaccine was the mumps and that was four years from start to finish most of the time it takes decades ai cannot speed up the slowest part of the trials it can't speed up the human testing but it can tell scientists where to laser in so that those vaccine candidates are much stronger it can look at drug interactions much more effectively it can show us what kind of toxicity it's going to have and all this hardware is accelerating simultaneously at the same time so recently there was a supercomputer out of oak ridge national lab which was the summit
supercomputer the fastest the second fastest supercomputer on earth and the reyes supercomputer both of them designed for deep learning 27 000 gpus again going back to the video game technology that is now changing artificial intelligence and converging with music and everything else and it crunched billions of combinations for for a week and came out with a brilliant analysis that united all of the different mysterious symptoms the brady kitten hypothesis is what it came up with and that allowed doctors to start looking at the covet as a vascular disease rather than just as a respiratory disease because they found that when it enters the nose and the lungs it binds to ace receptors which are all over your lungs and in your heart and in your your your kidneys but it also hijacks the body and tricks the body into up regulating those ace receptors in places where it's not usually expressed like the lungs and so more and more that's why they started to see people with high blood pressure and heart disease you know
tend to have much worse cases so these were the insights the feature extraction that the the machine was able to do and give scientists a new place to attack and what's even more interesting from that standpoint is that they're also identified nine other drugs that were already on the market that could potentially have efficacy so more and more we're going to see the the artificial intelligence tell us where to laser in so that scientists have a better time to market for all of these drugs it's also we're going to see more and more of the intelligence of artificial intelligence seep into our applications itself you're going to ask your phone or your glasses or your house ai what to do first before you go to the doctor and this is going to revolutionize medicine it's going to get medicine out to the farthest reaches of the countryside it's right now when you live in the countryside you may have access to a good gp but probably not to a great specialist so what do you
do you wait you generally wait you're too busy you have other things to do you'll get to it the next day but waiting can be fatal waiting is the very thing that you shouldn't do but if you can just take your phone and there's a little pin or there you take your your watch and there's a little pink prick in there that's constantly monitoring your blood pressure and triglycerides and it warns you you know weeks in advance to go to the doctor because you might have a heart attack before you go face down on your oatmeal you have a much better chance of surviving there were some researchers at boston university that were and were able to predict people with heart disease and diabetes with an 82 or 83 accuracy whether they would need care within a single year would they need hospital care within a year understanding that ahead of time means you can start to get the treatment that you need the problem is too many people get the treatment that they need
too late it's like piling sandbags against the flood it's already too far gone and that's what drives up the cost that's what makes survival rates low but now if you're able to look at your phone point it at a little spot on your skin and say should i worry about this only a fool would wait you'll drive to the next town to get the specialist and you'll begin to trust the algorithm more and more young people already do trust their algorithms more than they trust people and that's only going to increase and doctors and nurses will trust the algorithm too right now when you call the doctor the doctor has no idea whether you're a hypochondriac or whether you actually have a real problem and you get put into the same queue as everyone else if you live in berlin the idea of an appointment is please come down here tomorrow at nine when it opens and wait here for three hours and maybe you'll see the doctor but if i'm able to zap the
results along to the doctor the doctor will now know for sure that i have a real problem i should be prioritized in the queue now again that allows us to speed up the delivery of excellent health care so that's really going to change how we track disease how we learn about disease how we get help and it's going to change everything in the not too distant future now progress is never a straight line and so of course there will be setbacks along the way we've seen things like ibm watson for healthcare flame out dramatically it tried to do too much too fast and the problem is of course the people will look at this and say oh it's never going to happen now but of course they're not looking at the engineers working on the problem going back to the drawing board figuring things out people tend to look at the world and see it today and they see it exactly as it is now and they think it's going to be the exact same
way this tomorrow very people are very not very good at seeing big patterns in life they're not very good at seeing the huge sweeps of changes over time and all these things are going to continue to get better we'll have these setbacks it's like a river that's winding not just a straight line we recently watched back to the future the other day back to the future too it didn't age as well as i thought um but and there were a bunch of inventions in there and from the the far away age of 2015 like flying cars and anti-aging drugs that haven't gotten to us yet and so i caution you to take 10 years with a grain of salt it's easy to say what will happen if you do a monte carlo analysis of the future and see everything playing out it's very difficult to pinpoint exactly when but the roots of everything that i've been talking about are here in the next 10 years and remember again humans are very inventive species you know
agricultural revolution 12 000 years scientific 400 years industrial 200 information age 50. you know welcome to the age of artificial intelligence and nobody knows where the roads of tomorrow are going to lead but where we're going we don't need roads thank you thanks i'll make my way up the stairs we will still need stairs i'm afraid yes we will i i was really um taken by uh what you said you're tired of darkness i thought that might be a good book title you've got one coming up yeah this this is really interesting so of course you have looked into the future but how far do you think have you looked into the future i mean when back to the future i think is 30 years the future it's 30 years in the future okay yeah yeah so that was 2015 which has already gone by and really in this talk i just wanted to look at what's what's starting to happen in the next 10 years and the roots of everything i was talking about
are going to be there whether it takes 10 or 20 or 30 doesn't matter because we're going to start to see all this come together right right it's like you think about things like big biotech now the coronavirus puts a lot of pressure on society right it's like a diamond right is forged in this in this huge pressure but so are societies the pfizer i think inked a deal for 1.9 billion dollars for 100 million doses pfizer and astrazeneca have ink deals for 300 million doses with the eu in india those are just the two companies they're inking them with all these others so it's going to be billions and billions of dollars now that these companies have made there's an infrastructure that's going to be built they're not just going to turn it off tomorrow they're going to go they're going to pour that into all kinds of r d and new kinds of drugs and discovery tomorrow and they're already hiring tons of data scientists to build new algorithms and it's going to
be a virtuous circle that accelerates our ability to do medicine better and better so even if it all doesn't come to play where you have a perfect ai doctor in your phone or your glasses or your house or you know we have tremendous success with this vaccine and it takes longer to do other ones the roots of all this are going to continue to play out over time and it will take place right thank you very much kai asks first of all he says amazing talk love to listen to your flow thanks and then he asks trying to come up with a smart question i ended up asking myself what you might think about the scenario of carbon chauvinism machine wreckers and other extremist conservative forces combating any kinds of innovation and in particular anything linked to ai i mean you're always going to have the forces of progress and the forces of not progress in society and it's a dynamic force and both are actually necessary right if you just have unfettered progress right
then you have change that happens too quickly and you have these huge disruptions in society they're incredibly painful and if you just have people who want to keep things exactly the same way all the time then there's no progress right and so more and more society has been about how things change and yes i think for instance big media could absolutely kill many of the changes that we're talking about in the arts it's definitely going to become a battleground right to talk about well who owns frank sinatra's voice right well if it's gerald's voice now right and so they might try to strangle it but the problem is that that's going to work against them in the long term because you can't stop progress so just like napster came along they tried to force cds on everybody and napster came along and said no we want them digitally and so eventually they had to adapt and they did now we have spotify and you can listen to any you know song that you want in
the universe and you're carrying it around in your pocket so it's a push pull between all these forces i worry sometimes about the regulators coming in and defining you know um protocols for artificial intelligence before artificial intelligence is even out of the womb so how do you design a law or a protocol for something that isn't fully baked yet right i think that they need to take more of a closed hands-off approach to it or you know take their time with it um but again these forces aren't necessary they're asking people to ask questions you know they're they're saying look there are there is damage that ai can do if it's sentencing algorithms i've talked about these things in the past or it's driving cars there are real life and death situations if it's deciding who goes to jail or who gets bail who gets a job who doesn't who gets promoted people should be looking at those things with a conservative eye they're going to make mistakes but it's going to be the interplay
of these two forces to give us the landscape in the future and that's a good thing it's an okay thing are you at all worried that big summit might have a virtual daniel jeffries and and just put that on stage and let him talk and feed him with the latest ar articles as long as he says something super interesting okay um you know i don't it's going to be hard to recreate dan jeffries in uh digital granted but technology is advancing fast it's it's you know it's really quick maybe they're scanning my you know brain you know stem and they're going to put me upload me into the machine i think you know digital creations all those sorts of things they're super sci-fi but eventually they're possible we're already studying all the with the connectome projects in the world we're already studying all the we're trying to map the brain in the same way we map the genome we're trying to make full living brains so eventually if you can map and simulate an entire
brain you can probably simulate some type of consciousness and then so you can probably simulate the way that Dan Jeffries speaks maybe you go look at all the things that i wrote or said publicly and you kind of reverse engineer you know a personality for dan jeffries you know would you like to talk to a virtual dan jeffries or would to be too much in one conversation what do you think yeah yeah sure why not i mean i like talking to myself anyway and i like hearing myself talk right maybe it would be a little different um going back to being tired of darkness and you said like i'm tired of darkness i only want to talk about the possibilities and the chances right is this is this there's a 2020 thing or is this a paradigm shift for you is this something maybe other people should follow suit maybe should we should we talk more about the possibilities without talking about the dark side of things i i think it's good to talk about
the dark side of things but but i find more and more people don't talk about it in an intelligent way right whenever you talk about artificial intelligence they're focused on hollywood tales of like terminators and it's like we don't we don't need super conscious machines to for ai to do dumb things i actually worry more about narrow intelligence controlled by selfish egotistical violent human beings than i worry about ai you know super intelligence taking over the world i figure it probably can't do worse than we already do now so but i i wish people would talk about the problems that are actually existing you know look at when you have algorithms deciding whether somebody should get a loan or not that should be looked at very closely you know when you have an algorithm deciding whether someone should get hired or their resume gets through that should be looked at very closely there should be proceed human in the loop procedures to look at these things until the algorithms get very very solid and we
shouldn't just hand over all the power to that so i do think we should talk about those things but more and more i'm just tired of talking about the darkness altogether right i feel like we there's you you turn on the news today and there's nothing but darkness and fear and things to worry about and everything's going wrong and and somewhere there's somebody with a phone with no phone somewhere sitting in a forest by a waterfall and who's like perfectly happy and wondering you know because he doesn't have a phone telling him to be afraid all the time right right so from my standpoint i do think it makes sense for us to focus in on all the wonderful things that we're doing like i said in my talk you know if you're if you're a data scientist now and the only thing you're being hired to do is check attrition rates at your company and make people click ads you know what there's more exciting stuff coming right right and if you don't know
that maybe you don't spend your time on your passions maybe you go you know what art's useless i really love art i want to learn how to do style transfer and you know teach computers to paint and and and learn how to recreate voices and create wonderful music but you know there's no money in that there will be money in that you know follow your passion do those exciting things and that's why i want to i want to inspire people more now to think about those things so that they spend more time on those passions because then one day they're going to call from someone and they're going to be very rich remember that jeffrey hinton you know everybody told him don't study this crap it's a waste of time it's never going to work there's one university in canada that was keeping deep learning research alive for a decade and and they told you know kids don't go into this field it's a complete waste of time and now all these people are making
you know millions of dollars hundreds of thousands of dollars because they stuck to their passion they stuck to what they believed in and so i want to encourage people to really go broaden their horizons especially in terms of artificial intelligence think about the positive things they can do focus on it because we're going to need you more than ever in the in the next 10 or 20 years great thank you and even if we manage to create a real good 100 percent virtual copy please do come back in person for the next president thank you thank you daniel jeffries everybody thank you