[Script Info]
Title: 
[Events]
Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Ok so good morning everyone.\NI'll just get started.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,My name is Shailesh and I give these talks\Nalmost every year so this is a very
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,deja-vu feeling for me. The only thing\Ndifferent this time is the stage is
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,slightly thinner. But great crowd.\NGreat list of talks so far.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,So, Daniel called me a couple of weeks\Nago and said, 'Why don't you give a
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,keynote again?' And I said, 'You know, I'm\Nrunning out of things to say now.'
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,I've given four talks at different forums\Nwith The Fifth Elephant and I wasn't so
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,sure what I want to talk about. So, then\None of these days I was talking to one of
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,my non-geek friends and he was very\Nexcited about what I do, so he said,
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,'What do you do?' and I, you know, it was\Non the phone and I started talking to him
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,about this, that and the other. And for\Nabout 45 minutes I was rambling
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,and this guy was very quiet. I didn't\Nrealise he wasn't a techie and I was
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,going on and on, and after 45 minutes\NI stopped and said,
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,'Are you still there? Are you listening?'\NAnd he said, 'Yeah, I'm listening.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Can you tell me what do you do again?'\N(audience laughs)
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,And then I realised, how do I summarise\Nthis in 2 words?
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,So then I told him, 'Hey, I'm building\Nthinking machines.'
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,And that's when he said, 'Why didn't you\Nsay that before.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,It was so easy to say that, right?'\NSo that's how the title came by,
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,and obviously we're not building\Nthinking machines but
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,what I'm going talk about is towards\Nthinking machines, right?
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,So, we have a long way to go. So I\Nadded the word 'towards' later.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,So what I'm gonna talk about is all over\Nthe place. I'm going to talk about
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,philosophy, science fiction. I'm going to\Ntalk about algorithms and
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,I'm going to talk about, you know, deep\Nlearning and how to think about things
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,beyond deep learning. Alright?\NAnd let me give you a perspective
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,and then we'll start. So I'll take\Nquestions at the end.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Start working this.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Alright, so, I ended my last year's talk\Non this quotation
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,So I thought I'll start on this quotation\Nthis time.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,So I like this quotation because it puts\Na lot of things into perspective of
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,what we're doing, how our civilisation got\Nhere, and where we're headed.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,So it says, "Our technology, our machines,\Nis part of our humanity.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,We created them to extend ourselves, and\Nthat is what is unique about human beings"
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,And if you look at chairs, and dogs, and\Nanimals, and cats
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,They don't create machines to extend\Nthemselves. They just have instincts
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,and they follow their instincts. Right,\Nthat's very unique
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,about human civilisation. We've created\NTaj Mahal, and space flights, and internet
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,And so we've come a very long way.\NSo if you think about the tools, right?
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,The cavemen had tools and now we have\Na completely robotic assembly line
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,with no humans and you could turn the\Nlights off and nothing will happen.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,The car would get ???, right? We've gone\Nfrom just on-road, bullock carts,
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,to massive amounts of transportation we\Ncan do now.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,If you look at our ability to look further in\Nthe space, again, since Galileo,
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,we've made a lot of progress, ???\Nhe's certainly a thousand years off our
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Pluto fly by. So now we're able to send\Nsatellites into space.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,If you look at the first computer we built\Nand where we are today, right?
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,We have a huge data centre, and really, if\Nyou look at the whole thing in perspective
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,we have made an enormous amount of\Nprogress in the last so many centuries.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,So if youlook just at the technical part,\Nthe IT kind of intelligent machines,
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,we're not talking about mixies? and other\Nthings, just look at what AI
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,and deep learning, this stuff, has\Nproduced. Today's machines can play chess.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,And there's no human on the planet who can\Nplay chess better than the machine.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,I want to take a pause and think about\Nwhere we are.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,There's no human on the planet who can\Nplay chess better than the machine.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,There's no human on the planet who can\Nplay Jeopardy better than the machine.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,And recently, Google came up with\Nautomatic cars, so the machine can
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,drive cars and record show, that this cars\Nare better than humans under rider?
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,conditions. And they have much less\Naccident rates, and all the accidents
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,happened because of other humans drivers.\NThey're not because of cars.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,And recently also saw how machines are\Nable to create pictures, right, so this is
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,one of the things that deep learning is\Ninternally doing.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,And now think about all this. Just think\Nabout where machines have gone today.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,How many things they can do which are\Nway beyond our imagination
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,that machines could have done.\NSo obviously there's a lot they've done.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,But can they do the following?\NWe would want to stress their limits
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,So one of the holy grails of AI is to have\Na machine have a conversation with
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,a human being. We all know the Turing test\Nand the repercussions of this will be huge
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,We could think about how we talk to the\Ninternet today. We carefully craft word
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,for word queries, right, and you know, we\Nallow the internet to make mistakes
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,We craft queries again, and we take the\Nsuggestions or not
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,We talk to the internet like we're talking\Nto a 3-year-old
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Now in the daily needs of massive data\Ncomputers, NLP? and all its deploying
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,staff, imagine how shameful it is to talk\Nto a computer like a 3-year-old.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,So it's got the capacity of thousands of\Npeople but it can't understand language.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,So we need to change that. Now imagine\Nbeyond keywords what can happen
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,We can do question answering, but how do\Nwe do question answering today?
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,We've created Yahoo Answers. We've created\NQuora, where people can type questions
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,We do a match between the questions\Nand the answers, and then
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,we again do retrieval. So not answering\Nquestions.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Now think about conversations.\NConversation is an even more complex thing
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,If it works out, what are the\Nrepercussions? I don't want to study
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,physics from my physics teacher. I want to\Nstudy from Einstein or Feynman.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,We already know all the language and\Nknowledge of these people.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Can we not have a persona or a person\NFeynman or Einstein,
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,and have a conversation with that person,\Nright? So just imagine the future
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,of what will happen if we're just able to\Nhave conversations with the machines.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,So there's a long way to go between\Nkeyword search and conversations.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Can we discover a cure for cancer?\NThere are a lot of diseases out there.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Now obviously there is a lot of research\Npharma companies are doing.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,There's a lot of new initiatives on how\Nto use the high end machine learning
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,in pharma research. But my contention is\NI believe that the cure for a lot of
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,diseases is already out there. In all the\Nmedical literature, if somebody could
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,actually read them, hold that knowledge in\Nthe brain, in RAM, and do interconnections
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,we should be able to find a lot of things.\NBut what is the problem?
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,A single human expert, even in one field\Ncannot keep up with that quest of
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,knowledge, right. We forget some things,\Nwe want to read certain papers.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,And therefore, it's the other problem. \NWe have too much knowledge
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,and our individual brains are not\Ncapable of forming those connections
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,in the - because we can't even read that\Nmany docs, right?
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,But machines could do it, the way, and\Nthen there's progress.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Can we not find cures or new medicine\Ntoo.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Can I crack the next IIT Entrance Exam?\NYou laughing today, but you never know.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Five years from now, what will happen?\NWe should hope that if Watson is
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,a test of intelligence, if Igloo is a test\Nof intelligence, could this not be
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,a test of intelligence.\NThe ability of AI system to be able to
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,actually solve an IIT paper and get a\Nrank 1.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,What about, can I search all the video\Nscenes, which only have a goal shot
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,in the football videos and nothing else.\NI don't want to watch the rest of it.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,A lot of balls going here and there.\NI just want to see the goal shots.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Today I cannot do that.\NCan my machines be intelligent enough
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,to vision part, to actually find, this is\Na goal, this is a goal, this is a goal -
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,the rest of it is something else.\NSo we can imagine the applications now.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,We were talking about sarcasm a lot and we\Nall understand sarcasm is a very hard
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,thing to do. And imagine if you could\Ndetect sarcasm, what else can you do?
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,You writing an email to your boss\NYou're angry, you've written
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,a sarcastic comment, and ? says,\N'Hey are you sure about this?'
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,In the heat of the moment, can\NI put it this way?
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,So, like, today we do attachments. Can you\Ndetect sarcasm and things like that.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,And to me the holy grail of AI is not\Nreally all these big things,
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,but a really simple thing. Can a machine\Nfind a joke funny?
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,Now there are a lot of - don't know if you\Nguys watch Star Trek - but data entry
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,300, 400 years from now, is an android who\Nis capable of all these other things.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,He's a great supercomputer in human form\Nbut he's still struggling with humans.
Dialogue: 0,9:59:59.99,9:59:59.99,Default,,0000,0000,0000,,That's how hard the problem is.