The HLF Portraits: Raj Reddy

Video thumbnail (Frame 0) Video thumbnail (Frame 1337) Video thumbnail (Frame 5832) Video thumbnail (Frame 7836) Video thumbnail (Frame 9835) Video thumbnail (Frame 11080) Video thumbnail (Frame 15085) Video thumbnail (Frame 17085) Video thumbnail (Frame 18579) Video thumbnail (Frame 23079) Video thumbnail (Frame 28585) Video thumbnail (Frame 31079) Video thumbnail (Frame 32828) Video thumbnail (Frame 36085) Video thumbnail (Frame 37329) Video thumbnail (Frame 40085) Video thumbnail (Frame 44580) Video thumbnail (Frame 45830) Video thumbnail (Frame 49080) Video thumbnail (Frame 50587) Video thumbnail (Frame 52079) Video thumbnail (Frame 58332) Video thumbnail (Frame 60579) Video thumbnail (Frame 72829) Video thumbnail (Frame 74080) Video thumbnail (Frame 75832) Video thumbnail (Frame 80081) Video thumbnail (Frame 81984)
Video in TIB AV-Portal: The HLF Portraits: Raj Reddy

Formal Metadata

Title
The HLF Portraits: Raj Reddy
Title of Series
Author
Contributors
License
No Open Access License:
German copyright law applies. This film may be used for your own use but it may not be distributed via the internet or passed on to external parties.
Identifiers
Publisher
Release Date
2018
Language
English

Content Metadata

Subject Area
Abstract
The Heidelberg Laureate Forum Foundation presents the HLF Portraits: Raj Reddy; ACM A.M. Turing Award, 1994 Recipients of the ACM A.M. Turing Award and the Abel Prize in discussion with Marc Pachter, Director Emeritus National Portrait Gallery, Smithsonian Institute, about their lives, their research, their careers and the circumstances that led to the awards. Video interviews produced for the Heidelberg Laureate Forum Foundation by the Berlin photographer Peter Badge. The opinions expressed in this video do not necessarily reflect the views of the Heidelberg Laureate Forum Foundation or any other person or associated institution involved in the making and distribution of the video. Background: The Heidelberg Laureate Forum Foundation (HLFF) annually organizes the Heidelberg Laureate Forum (HLF), which is a networking event for mathematicians and computer scientists from all over the world. The HLFF was established and is funded by the German foundation the Klaus Tschira Stiftung (KTS), which promotes natural sciences, mathematics and computer science. The HLF is strongly supported by the award-granting institutions, the Association for Computing Machinery (ACM: ACM A.M. Turing Award, ACM Prize in Computing), the International Mathematical Union (IMU: Fields Medal, Nevanlinna Prize), and the Norwegian Academy of Science and Letters (DNVA: Abel Prize). The Scientific Partners of the HLFF are the Heidelberg Institute for Theoretical Studies (HITS) and Heidelberg University.

Related Material

Turing test Multiplication sign Internet forum Musical ensemble
MIDI Electric generator Observational study Multiplication sign Mereology Expected value Tablet computer Category of being Elementary arithmetic Personal digital assistant Order (biology) Universe (mathematics) Video game Right angle Family
Control flow Energy level Right angle Family Field (computer science)
Point (geometry) Category of being Standard deviation Goodness of fit Term (mathematics) Direction (geometry) Universe (mathematics) Right angle Student's t-test Ranking Figurate number Resultant
Mathematics Direction (geometry) Physicalism Student's t-test Neuroinformatik
Point (geometry) Satellite Group action Inheritance (object-oriented programming) Direction (geometry) Gender Multiplication sign View (database) Student's t-test Power (physics) Number Process (computing) Term (mathematics) Universe (mathematics) Video game Right angle Ranking Position operator Physical system Social class
Principal ideal Uniform resource locator Different (Kate Ryan album) State of matter Number
Process (computing) State of matter Video game Student's t-test
Point (geometry) Open source Civil engineering Virtual machine Student's t-test Disk read-and-write head Turing-Maschine Computer programming Theory Machine vision 2 (number) Neuroinformatik Goodness of fit Mathematics Semiconductor memory Matrix (mathematics) Computer engineering Punched card Computer architecture Social class Programming language Multiplication Electric generator Electronic mailing list Computer scientist Degree (graph theory) Type theory Computer science Video game Quicksort Row (database)
Point (geometry) Degree (graph theory) Context awareness Mathematics Stapeldatei Direction (geometry) Computer science Public key certificate
Symbolic computation Turing test Pattern recognition Multiplication sign Sound effect Computer programming Formal language 2 (number) Data management Word McCarthy, John Speech synthesis Right angle Data conversion Local ring Reading (process) Physical system
Point (geometry) Word Lemma (mathematics) Speech synthesis Right angle Neuroinformatik
Moore's law Algorithm Dialect Different (Kate Ryan album) Multiplication sign Virtual machine Speech synthesis Independence (probability theory) Computer Mereology Formal language Physical system
Context awareness Touchscreen Software developer Keyboard shortcut Virtual machine Interactive television Similarity (geometry) Plastikkarte Disk read-and-write head System call Symbol table Neuroinformatik Word Process (computing) Computer configuration Right angle Smartphone Fundamental theorem of algebra Exception handling
Group action Artificial neural network Multiplication sign Reflection (mathematics) Range (statistics) Computer music Disk read-and-write head Mereology Machine vision Field (computer science) Computer programming Information technology consulting Formal language Neuroinformatik Moore's law Process (computing) Robotics Personal digital assistant Speech synthesis Right angle Musical ensemble Task (computing)
Dependent and independent variables Group action Hill differential equation Neuroinformatik Spacetime
Dean number Mathematics Multiplication sign Rule of inference Exception handling
Turing test Point (geometry) Computer chess Multiplication sign Mass Student's t-test Mereology Rule of inference Neuroinformatik Hypothesis Formal language Semiconductor memory Different (Kate Ryan album) Theorem Physical system Task (computing) Programming language Regulator gene Artificial neural network Expert system Computer scientist Computer Proof theory Category of being Word Data management Integrated development environment Computer science Right angle Game theory Arithmetic progression Spectrum (functional analysis)
Group action Building Multiplication sign System administrator Zoom lens Source code Real-time operating system Disk read-and-write head Semantics (computer science) Computer programming Neuroinformatik Medical imaging Different (Kate Ryan album) Single-precision floating-point format Circle Endliche Modelltheorie Physical system Pattern recognition Programming paradigm Moment (mathematics) Constructor (object-oriented programming) Lattice (order) Sequence Connected space Dean number Type theory Right angle Arithmetic progression Arc (geometry) Point (geometry) Motion capture Student's t-test Machine vision 2 (number) Frequency Goodness of fit Robotics Term (mathematics) Associative property Traffic reporting YouTube Task (computing) Knowledge-based systems Faculty (division) Word Loop (music) Software Universe (mathematics) Formal grammar Speech synthesis
Crash (computing) Overhead (computing) Robotics Cellular automaton Website
Knowledge-based systems Expert system Knowledge engineering Planning Rule of inference Computer programming Neuroinformatik Category of being Operator (mathematics) Telecommunication Logic programming Speech synthesis Endliche Modelltheorie Physical system
Multiplication Multiplication sign Internet forum Telebanking Extension (kinesiology) Metropolitan area network Neuroinformatik
[Music] fess are ready we begin at the beginning locate us in your childhood where our own I was born in a little village in
India it's still a small village there are about 500,000 people now there used to be 500 and at that time this was in
the 30s just before the war and I didn't know for anything about the war and he was I was growing up it was an agricultural community my father had lands and we used to kind of be there
you know for a night first ten years I never left the village so expectations for you and your life would be as a farmer no that's an interesting part there were seven children four brothers and three sisters and my father decided unilaterally that the first two of them were going to look after all the properties and the other two can go do whatever they wanted and you were not a first time I was the third dismissed and it turned out it's like you know when you read about Dukes and barons in in England where you feel
the firstborn you have to be there and if you're a second then you sent after the army or something yes and the third becomes a pastor yes well the friend order for those things in my case it was just I was the third child and he said okay whatever comes we can do so you you weren't restricted in the expectations from were their hopes for you or yeah basically when they wanted to threaten me with dire and the possibilities my father would say see unless you study well you will be like that other guy who is a shepherd going and behind the shine near the ship and if that's what you want to do okay be by my be my guess it and it I'm not sure it worked I was anyway motivated but studious no there was not studious okay I was basically I didn't want to study I wanted to have fun well most of us do and so throughout you know until my undergraduate yeah until I graduated I would kind of have fun for 10 months in a year and study for one month that were the exams for the exam study for one month pass the exam so that if I get all about it before we get you to university what kind of elementary secondary education did you have yeah basically the first five years I was in first five of schooling I was in the village school right and in those days this is late 13 early mid 40s the UN paper and pencils were not easily available even a slate and things lever so then we learned how to write letters by writing on the sand and you know and it's not just me it turns out there's a whole generation of people in India that learned how to write by writing on this hand so and then I had to leave the village to go to
middle school and high school and so on right so in those kind there were no hostels so he kind of stayed as a paying guest and usually somebody you know kind of fad took you in and then you studied and you're all break out by yourself so from the age of 10 till now I've always
been away from home and what about
teachers is the level of teaching once
you get to the secondary level would you say good good the teachers were good you know what I mean is they did what they were supposed
to do but they did not turn you on basically they did not and ER tried to make it you know make you into geologists or something like look at his rock you know but that kind of thing happens here in the you know in my grandson goes to brown and out of the blue he came back and said I've changed my field from biology to geology I said why because the teacher was kind of exciting and you know showed him all the kinds of things okay I'm gonna have to
get you motivated at some point your father says do what you want the teachers are not pushing you in any direction I'm guessing you're an okay student in terms of results maybe yeah what was a good student you were mastered I was not the top student right when I was among the top 10 percent or top 5 percent or something most people were not necessarily going to do university but you yeah that's what I my father decided since I'm I don't have to be there to take care of the property and so on if I can get into a good college you would st. pay for it he would support you and so when I graduated from school I graduated in the
top ranks and then he said okay and if you can get into a good school I'm not going to stop you and so I ended up in a place called Loyola College in Madras those days okay is no culture nation and Loyola College you know as you might expect a Jesuit school right very good and probably high standards high standards but again rote learning and basically in India the learning was not to teach you to think they teach you to remember facts and figures and things
like that and so you know I was okay with it but that you know these days I tell the students and you know what we
need to teach people is learning to learn and learning to think and learning to live none of those are there in any curriculum any the world you need to teach computers that - did I have to learn that's the way in a using computer set to learning to learn is probably the fundamental skill we need to have but that didn't happen there what kind of direction did you have a major you must have yeah I was I was good in maths but I was only math and sciences and so in Loyola College I did a math major mattheum they what they call an MPC math physics in chemistry major that kind of
automatically kind of move you in the direction of engineering rather than medicine right and so after I finished Loyola College I went to one of the oldest engineering colleges in the world there this year they're having the two hundred and twenty-fifth anniversary and I'm going to be doing this topic in July and probably one of their great graduates yeah one of the great ok fair that so and I was admitted to that let's go and there's a very interesting thing and these days I talk about there's a problem with the way we admit students into Carnegie Mellon or Harvard or anywhere else namely if you are the best in the national rankings you're admitted by SAT scores or something but those marks are a function of whether your parents are educated whether you had good school we went to the top schools and all kinds of things you fear a child in it from a ghetto and your teachers are not that good even if you're basically pretty good you don't get in right right and so we admit the best educated students not the best students yes then so there is a book called the tyranny of meritocracy or something which kind of makes a point about this but in general California system did a great job they in 2002 they had Supreme Court of California said affirmative action is unconstitutional so the Regents to their credit said okay then we're not going to use just SAT as a way of admission we are going to admit people if they're a top student or top 4% in their school and so if you come from a school in the ghetto there's going to be a top 4% exactly right exactly so that whole idea of what we call geographic equity and gender equity the other thing is when you get to beyond about tenth class girls drop out many girls right you know if you go to any university they're not that many you know as many boys there's a lot more boys than girls right and that's worse in India but here for example at SIA moved this year because we've been
trying to kind of move in this direction for the first time the number of girls entering CMU is more than boys very first time this year I'm talking to the Falls for about Frank education right the 20 year old you might not have had this view of life I want to still meet the term here so basically I want to kind of bring up the consciousness of the 20 year old when they are Eureka I want you as the 20 year old yeah when they are in a position of power they should say how can we deal with this problem of meritocracy raise you put you
back in you know in my own trade frontier all yes I was not even thinking about any of this any of this but it so happened the admission was based on geographic equity in here in college and so I got in so otherwise principal yeah there was a principal where they said we have you know 200 engineering seeds we're going to admit them there are about 20 districts different populations we're going to assign some number of seats for each district you got in through geographical advantage yeah and geographical advantage in two ways you
know turned out I come from a location ninja where I am in the corner of two different states we had lands in both
yeah yeah it's like I wanted to go to Chennai I could use Landsat maybe have that state you know I rather pray otherwise I could were Andhra Pradesh so so I'm gonna be your father now you're in college I'm paying for your education yeah you're an engineering school now what are you gonna do with your life you didn't ask that he didn't
ask them now I didn't ask either we just said you know it is known globally yes that if you went to college and we get to look in the engineering college then you must be a good student and you your life get a job you get a job you get a job when do you catch fire because you do that I employ women that happen in Australia okay so I graduated okay immediately there was an internship
opportunity in Australia for Commonwealth students oh yeah so I said oh okay why don't we go that we had nothing else to do and three of us applied and went to Australia where we landed in Melbourne and and so I was an intern and I spent a year as an intern I was a civil engineer yes nothing to do with engineer and this was in the 50s yes so I you know after one year of internship I said what am I going to do I said let me join you know get a masters degree in Osterley or something so I joined University of New South Wales okay and the first day I was there is when I caught fire what happened was the head of the department my adviser and Stan Hall said hey I'm going to go to the computer center to write a program on the computer why don't you come I said sure I said what is the computer and what in those days a computer could mean somebody who computes using a machining and so he had just come back and from a sabbatical in England where he was using these computers it's called English Electric juice mark - that was a kind of a second-generation Turing machine well children designed a thing called ace Acera computing engine or something and this was deuce the second generation of ace okay and it was it only had one kilobyte of memory and that's mercury delay line memory so the main memory was carved punch card so you did some read some data and did some computing punched it out and then read it back in and did some more computing and as it's exactly you know what the the punch you know the looms you know are loose used and you loved it or you liked it or you were so basically he as he was writing the program he kind of thought aloud I said I'm going to do this and put you know it was all a matrix operation and we take this matrix and multiply by and I was observing it and it didn't take me very long to figure out I said I can do that for you from now onwards you don't have to come in the evening so that's I became his program well and that and so and after a year of I finished my degree and I said I don't want to be a civil engineer it looks like I'm doing well here and it turned out in those days in 59 there were no computer scientists anyway anybody that has even been exposed to computers so IBM in Australia was looking for someone who had some experience and I was it in and I got and I had you know probably the best
education of in my life at IBM so what happens was IBM had this you know idea saying anybody you are likely to hire really have a liberal arts education from Harvard or someplace they won't know anything about computers therefore we have to train them from the and then give them and so it's kind of a you get a two-week course and then you go practice for a month and then come back for two more weeks and so on so it's I spent half of my three years in attending classes this is like almost like another degree and but but the idea of learning by doing you learn how to do something and then immediately apply it well a partnership it's more than an apprentice in apprenticeship like bricklaying yes you're observing and then here you needed to understand the theory that to two weeks was to kind of understand the architecture of the computer and the programming language and all the other things and you did some type programs then immediately you came and they give you saying okay go to this customer and do this program the next stage of catching fire I would think would be when you began to have a vision of where this might go are you at this point thinking of the future of the computer so I I was I'm not sure I was thinking about the future of the computer but I read articles by Newell and Simon who were here from Australia and Minsky and McCarthy they were talking about this artificial intelligence I said maybe I should you know if I'm going to study you know that looks like something I might enjoy you know I had no clue what it was this was in 59 60 and so I was reading papers I was kind of you know also doing some postgraduate work with the same teacher that I because he said yeah now that you're almost done you can probably get a PhD but I didn't want a PhD in civil engineering so I applied to Stanford and I applied to Carnegie Mellon only to schools because that's the way well tell me why Stanford and Carnegie Mellon seemed like the good next stage because that's where the AI expertise was Lowell and Simon were here McCarthy was there okay and I was not admitted into CMU they said you're on a waiting list or something Stanford admitted me so I said I went to Stanford you must have written some sort of proposals to what you would like to study there yeah basically they had a computer science program in mathematics department okay I applied to that program it was a second master's degree in computer science and I'm not sure if I wrote an essay or anything but I might have we know but my previous record was all kind of straight-a in University of New South Wales so they have raised on that record they said okay we're probably better prepared than many people to apply I said I also him besides that I also have three years of experience working with a behavior exactly based on that I think they admitted me because in the end this is a big question you can answer it anyway it seems right but it seems to me for people with the kind of competence that
you were building and of course developed immensely there's a point where you were saying I'm going to go in the commercial context or I'm gonna do
an academic direction so yep so I'm not sure I explicitly thought about it I
could have but after after I finished my PhD before we'll go to the PhD I need to say that was the birth of so-called computer science that I was there from 1963 in 65 the department of computer science was formed we were present at the creation he I was present at the creation and because I was already well into the PhD I finished my degree in 66 one year after the department I was the one the first PhD in computer science from Stanford Wow which would make you the first probably well not in the world but yeah so the the it's one of those theoretical things there are a lot of other PhDs that got a PhD in math department or Business School or something they were computer science PhDs but they didn't have in their certificate PhD in computer science right they were probably the first batch here and there and I was one of the first again I keep thinking now no longer of your father but of your mentor now who is I suppose but Karthi tonight what is he telling you if he is about your future what will
tell me what your PhD was on my PA you know basically I was taking a course from John McCarthy and on symbolic computation and he said hey by the way we just got a PDP one yet delivered it has an A to D converter if anyone want things you might be able to you know recognize speech and I said sure I'll go work on it and I thought it was the beginning within a few three months I had the system working and doing what we call wallet recognition if you said I it said okay he said ah he's an e so and that's the first time I also got my shock they we were funded McCarthy was funded by DARPA at that time yes and so now I won Sutherland another Turing Award winner you might have interviewed already was one of the DARPA program managers he comes into Stanford and Brian McCarthy saying hey he needs an exciting thing literally yank you know I was not a young kid I was already 26 or something ROG is doing and the young PhD
still but a young PhD sure and I was not yet a PhD I was working yeah going so Ivan came and said said various vowels and he'd recognize day then he whistled him to it and he'd said you said E so that's the first time I realize the speech recognition is going to be very difficult because it is not enough to simply know what I said it's also important to know what I did not say yeah so this is essentially that I tell this to a story to people if I don't understand the vocabulary if I can take me ru and put it in a Russian classroom yes I don't understand the words I don't have to understand anything I can simply say well I'm gonna give you an exam and you failed I may be the brightest kid right but if I don't understand what you're saying I am out of luck and the same thing happens to kids who are second main languages local language right and they're going into this English medium school in India when you go to college everything is English medium suddenly they're thrust into this situation where they have no clue I happened to me when I went to Loyola College I went from a Telugu medium school and certainly for six months I couldn't understand a word I could understand with the words but I didn't said what they're saying fortunately I could read books textbooks understand them and to the effect means as I survived but in general this is the same situation we find now but we're just about getting to a situation where a I will help us to solve the problem
we'll come to that later yes but what's your already launched you can't get a PhD and in in this problem design lemma
of speech right what is the insight you've come to that the PhD delivers at that point all that I said was it would be great if we can build computers that can understand speech even that anticipation was a new idea probably right and the interesting thing was we all thought you know maybe we'll solve it in 5-10 years in that 50 years will not yet there and you know we understand lots of things much better and but as I said we are on the words basically the
latest machine learning algorithm deep learning and so on are able to recognize spoken language in any language provided it has been given enough data like a million hours of speech and we're now
capable of having machines ya know not only and these things could not have been done even 10 years ago you needed computing power that was like a million or even a billion times more than what I had when I was working on it in the 60s so why aren't you discouraged no no I I knew that we will be able to do it there's a question was I said you know if you talk to me maybe in 1999 and 2000 I said it may not happen in my lifetime kind of computers recognizing unrehearsed spontaneous speech from open population we've already built systems which would recognize my voice or your voice with independent dictation was already available from Dragon Dictate and nuance and various other people but unrehearsed spontaneous which is very different I say people speak you know and then on top of that it's I know it's not just one person it's any accent from any part of the world speaking English and now take it to Chinese you know although there's one Chinese language most of the people speak there are literally 40 or 100 different dialect dialects of course and in India there are 22 official languages and you have to learn all of them so
explain to me the aspiration to do this why isn't it enough to have interaction with machines yeah that's the fundamental question I'm glad you asked it turns out in the world 3 billion people about 40 percent are either ilnur traitor similar trick they can they can even read some symbols words but they don't understand what the sentence means yes and so right now that 40% are out of luck they are not benefiting from our technology and as I was saying live in another world because they and they are all of them are big eaters smartphones they can actually you know use it to call somebody and talk but but they're not benefiting from the whole thing so what is what we are now doing is not we you know large companies given the large enough data from all these populations you don't have to know how to read or write to benefit from all the things for example if I had a smartphone I can say computer read me the newspaper today and to show me the pictures and things and then I can say what is the headline you know and then it'll treat the headline okay now skip to the next one you know you can kind of essentially read the newspaper that is it's what we call voice computing you imagine a world where you don't have a keyboard you don't have a screen just like people when I want something done by Vivian I say get me you know Peter right she even knows the context it must be Peter back from before I'm not supposed to talk to him today see it's the background to understand all that you know is there and so that's what is voice computing so it turns out Amazon echo finally last year came up with a device which is just six there doesn't I think that is completely non-intrusive Sydney was doing the same thing on iPhones except they're intrusive I had to press and hold some button yes I said now I'm never going to do that so it was it is possible finally you know it took Amazon to demonstrate it now smart speakers that are called everybody has a smart speaker okay I'm going to give you credit for some of this development but I have to get you from your PhD right to your next stage of intellectual development so I so basically at that when I finish the PhD I had an option I didn't have any option I said I didn't try very hard either I went to the department head at Stanford George Forsyth I said George you know
I'm looking for a job you do you think you might be able to use me and it turned out there were not many computer
science PhDs he was desperately looking for some so two of us graduated he hired both of us building human and I hmm and we stayed there for three years a you know bill left after two years because you know other places were desperate and he was offered a head of the department job with tenure what was your task at those three years in those three years
after your PhD what were you working on same thing so basically I was part of the AI group of McCarthy okay artificial intelligence group and there were a lot of things we were working on in those days essentially everything we're working on today we were working on robotics and we were working on language
we weren't under working on speech we were working on computer vision we were working on various other things like music computer music music synthesis was done there so there's a talk why me don't mind you'll find it about early reflections on the early days at Stanford which kind of explains everything we're doing today we were doing already in 60s except now we have million or even a billion times more computing power that's deities that makes it different it's not our ideas are exactly the same as before they are they often say about knowledge in general at least in the West yeah then in the 18th century an intelligent person could know everything right broadly by now know so in a way also you have this advantage in the computer world that early on you could be involved in a whole range of things probably a young scientists that would not have this out Frank exactly you hit the nail on the head basically in those days it was like going and mining for gold in a gold field everything in a rock you pick up his gold okay every problem you could think of and and you could make serious progress and get a PhD and that's not no longer the case so you need to kind of learn all that stuff right so in 60s we had that advantage and so people who are looking working on all kinds of things including synthesizing music and there was a one person can call be the MD PhD I was working on miracle consultation for like a psychiatrist he built a program called Perry so if you wanted to train doctors and how to recognize paranoid patients
and you know the doctors could train themselves using this computer and no matter what you said it will kind of give you a paranoid response saying why are you doing X Y Z to me right so I'm curious maybe this is not a very important question but the intellectual climate in the Bay Area in general when you're there because Berkeley is also yeah developing groups of people
interested in this it must have been a very exciting community yes it was but vertically was not yet kind of at the pinnacle but there was SR I collateral Research Institute and on Stanford and Stanford AI Labs was kind of off off in the hill somewhere because their interest of space for them right here was growing you know and there was a whole you know Hewlett Packard was there and all the silicon there early
seeds of Silicon Valley Raytheon and a whole bunch of other companies were there why did you leave this intellectual paradise interesting I know obviously I was after three years it was time for them to decide you know what to do with me the department promoted me to associate professor yes so it went to the Dean of the school the Royden from math department city you know we have this rule in Stanford where we don't hire our own PhDs I already made an exception for Raj three years ago I
think he should just go for one year to somewhere else and can come back okay so I could have gone to Berkeley but I wanted to you know so I came to CMU for one year and stayed here for 50 years right right so the idea was to go back to Stanford but you never left
yeah what did CMU what was its emphasis at the time you came I came because of New Orleans salmon two of the four founding fathers of artificial intelligence were here and there was also another person very well known computer scientist Alan Perlis he is a giant in programming languages he designed many languages and the three of them together this was the place to go if for people that understood right because most people didn't understand they would simply apply to some place like Harvard to go into computer science there was nobody there and till much later when I won Sutherland joined Harvard right and so it's it's an interesting what is the year now that you to CMU we I came here in 1969 49 years ago 50 years ago right what what is the stage of AI at this point basically in 1969 we had a challenge you know of what can a I do and Lowell and Simon had built a six six by six chessboard because there was not enough memory you had to claim eight by eight and and had a chess game working they had a symbolic computational machine which could prove theorems and various other things that I know and everything and they were beginning to study for at that point artificial intelligence meant computers that can demonstrate some kind of a task that would normally be considered intelligent like if you play chess yes and if you played you know with food theorems and in general problem-solving was the name that we were they were using and they worked on for ten years human problem solving psychology because they were also psychologists so they were trying to model human mind so to speak is this related to expert the expert systems systems that will come later yeah basically one of the students that from CMU was at Feigenbaum yes Edie went from here to Berkeley and from there came to stand over in 65 and he was on my thesis committee was he know because you will got the Turing award one day together so this is very significant right and so you know I I was I was always kind of new add and close you know in the sense but he was working on certain things you at the understand furred AI lab around the concept of capturing expertise he was working with biologists and and you know repeat mass spectral experts Lysol I spoke with him yes right and so I that was there there so he was saying how can can I make a computer behave like a PhD and so that's that's where he spent you know last the next 20 years on made a huge progress and I think what you've also at least to my ears told us is that I'm not saying we're stuck at this stage but we are still trying to get computers not to develop some new ways of doing something but to replicate human beings yeah basically we were asking the question not what computers cannot do so to ask the question can we demonstrate a computer doing something which when done by a human being would be considered you know interject so that was artificial intelligent was just replicating human you not even replicating human that's a bigger word do things that commanding to human
behavior human being problem solved problem so at that time they were mainly worrying about things like puzzles and nolan simon spent a lot of time looking at how human being solved puzzles and then tried to get the computers to do the same thing so they've got this young managers come from stanford where does he fit into this so basically i i they said you represent a different part of AI we're not doing it this is perceptions computers that can see hear and speak and write and walk right robotic you know why don't you come and do that here so you would a specialist in a way yeah here this is they wanted someone complementary to whatever they were doing because that would make it you know what did they do to keep you you were gonna come for one year what happened now basically this is was always an empowering environment that is he wanted to do something you could do it there was no rules and regulations and said no you can't do that you have to sign this paper and that paper there were rules but you know but in general within computers in computer science department things were very inner empowering you know you could do your
thing and nobody got in your way and then I went to an Alan Perlis with the head of the department is it Alan I've been here for six months you've not even had a single faculty meeting and he said what faculty meeting I don't believe in faculty meetings he said I believe in hiring the best people and leaving them alone to do you their work and I will take care of all the administrators well Sarah dies another yeah because almost any other department even here would have faculty meetings yes they had to kind of find I have a discussion and debate and whether we do this at that or something else and for me I did want to attend faculty meetings yeah and later on when I was the Dean I didn't want to attend the president's council meetings and so I went to one meeting and I said I don't need to go to this so I appointed asked my associate dean to attended the president was not a happy camper this is the important meeting I said you know his name was and I said if you have a question you need my opinion on I'd be happy to do it but the things you I most of the things you talked about President's Council meeting I have nothing to contribute therefore it's a waste of my time based this was like a three hour meeting once a carnegie mellon produced an anarchist it would be an arc but the the empowering concept comes from the image it's called a reasonable person principle okay not everything was written down but if somebody wanted to do thing they could do it and it would be tolerated if not you know Anchorage at least tolerated as long it is something a reasonable person would have done an anarchist is not a reasonable third it turns out we had a prime situation of an anarchist was a faculty member in the finance department whose main art was anarchy he wanted to burn down the computer centers were down the bill very concept yes and and so and when the promotions came in I was in the I was already the Dean at that time I said what the hell why are we promoting this guy who wants to burn our computer what is art in there you know I want to you're not intellectually lonely at this point because you have this fascinating group with you yeah but in terms of your specialty yeah you are intellectually lonely I'm good at asou it turned out fortunately four or five of my students from Stanford also came with me really yeah and then you know there was a period in time I probably had 1520 PhD students at that time was this built as a sub department how what was the construction at that time there were just three full professors and four associate professors I was one of the four okay and the whole department was like 10 at fault people this small department and then you know we would have admit like 20 students every year so they'll be there here so there may be a hundred pH the only ph.d program no masters no may be romantic of me to put it the following way but when is your next Eureka moment the Eureka moment after that happened when we had DARPA you know challenge so in 1931 Allen Lowell chair the committee on a speech understanding systems report and that became a blueprint for five years for DARPA and say they're funded MIT Lincoln lab as our eyes CMU and so on about five different centers all with the goal of building a thousand-word will camera with speech recognition system connected speech circle and we you know everybody built it it turned out in 1976 we had a run off and we had two systems here saying harpy and both systems in fact in their YouTube videos and about both of them you can see you know at that time all the rest of the systems did do connected speech recognition but it could take them an hour to respond take a sentence and take an hour and then respond but whereas both of these systems are pre system was essentially real-time and here's a system was took one or two minutes and both of them you know used slightly different ARP was using hidden Markov models that was kind of pioneered by Jim Baker one of my students who went on to build dragon systems and and the other one the hearsay system was a rule-based knowledge based system like other knowledge-based systems of that year that L in the 70s and so we decided there are different kinds of knowledge we have to capture knowledge about Cylons knowledge about words knowledge about grammar knowledge what semantics if I said pick you know pawn to King four and if we listen to the voice and then make the move right that that's all semantically correct that would could be so all these knowledge sources constrain what you look at if you don't you know it's like any other puzzle it could be any sequence right ba-ba-ba-ba-ba-ba heesu could be a sequence except not correct construction in English right and so so that kind of though the whole purpose of knowledge sources was to constrain the search otherwise it would have been you know million times more flex and taken million times more yes and then by using this hidden Markov model integrated network we were able to kind of zoom in and the answer very fast when de and this builds that was the Eureka moment in 1976 yes when we all had demonstrations and everybody demonstrated the connected speed thousand word vocabulary systems except there and that report it didn't say you have to do it in real time because we all understood but what they did say is assume there will be computers which are thousand times more powerful than what you have and and if you can make it work you know or no more than a thousand times real time you know if it's that a sentence of one second and you know you should be able to recognize it in thousand seconds it's wonderful guidance yeah so it was in and so it was a very you know very powerful paradigm because all of us would demonstrate something would all be together and then everybody had to explain what they did to the whole group so if I did very well by in two months they could all implement what I did it was a very effective loop on a research progress in some sense because they don't have so much time you've been involved with so many intellectual tasks administrative - we won't talk about the administrator with the intellectual task and I want to get to robotics basically besides speech I had some students who are doing computer vision face recognition style right we had some students doing robotics type and then about 1979 ten years after I got here I was in a meeting with the president of the university excite and Alan Lal and
so on they said why don't we have a robotics Institute why don't why don't we do a robotics research I said we are it's just that it's just me and a few students and if you want to have a major impact we need to set up a robotic Institute and I need a million dollars from you immediate and to their credit they they gave me a million dollars and excited said this is not a grand this is a loan with interest we are to pay it back to
the University and that's a since then I borrowed money for many of the fact including that building there we borrowed five million that we didn't have to finish it up and then I had to pay it back with interest the way you know from with from the overhead very motivating no I was not more difficult
I'm right there I got an interest-free unsecured loan
what are they going to do I already had tell you they can't fire me right yes sir so you set up a cellar from robot robotics Institute Institute and then
already we knew lots of things we wanted to do we wanted to do lights-out manufacturing a fully autonomous factory or at least a cell and we wanted to have autonomous vehicles autonomous see vehicles land vehicles and air vehicles and the current cars that drive themselves came out of that idea the drones came out of that idea that was nurtured here yeah in fact we had a autonomous helicopter that was used for Three Mile Island Brno they're not the Mellen when the 9/11 thing happen is we sent the autonomous helicopter to this crash site in Pennsylvania and because we they want you to know what the hell was going on nobody could physically go there and see because it was smoking can you explain briefly to a layman what
the what needed to be solved to create that a timeless so basically first thing you need to understand is all of this comes from the basic idea that computers can do what human beings can do if you can become a human being can drive a car computer builder if human being can you know never get a plane computers available then the question is what does that mean yes how do you do it for the first 30 years we were kind of hoping that we could have knowledge encoded into rules rule-based systems and knowledge-based systems that's the thing that Feigenbaum championed and expert systems are that all of that kind we still use them except you know what has happened is it's like the telephone system if you remember the when the first telephones came there are human telephone operator yes that would plug that operating right and and it was said if you did not invent it in a switch a mechanical switch our electromechanical switch and electronic switch 90% of the population of the world will be switchboard operators connecting the other 10% so the same thing would have happened here we don't have enough manpower to write all the programs needed and all the knowledge engineering needed to capture the knowledge to make this systems be intelligent and so very quickly it was clear it took us 20 years to learn the hard way that we needed systems that can learn from experience so the main emphasis these days in the last 20 years is systems that can learn from experience we built the first speech system the harpy system and so learn from experience using the hidden Markov model learning of just the phonemes but now is much more sophisticated much more sophisticated so the main lesson is we do not have enough people to capture all the knowledge and put it into the computer the computer has to do it by itself so for example if - when you push this to the extreme I can imagine a day when there will be no probe computers will write their own programs they will look over the shoulder of what you're doing and I'm doing and then do the same thing for example and we are already working on things like computers
that will pay my bills yes I will not I won't know how to really progressed but what it will do is it will look over my shoulder and see what I do when I go to
the bank online banking and pay the bill and the next time it says oh you're trying to pay that bill I already paid it so should I be terrified at that future rated you know all that is doing is saving you some time why should you be terrified so the issue is people then make extension of that automatically saying this is going to replace me not going to happen the the concept of an AI is never been about human replacement it's about augmenting human capabilities enhancing the mental cables anything you do with your man when it was your brain it was your mind computers can help you do it better faster cheaper so you can do it 10 times faster you may be able to do it a million times faster sometimes like multiplication and things like that but it's all that it's doing is relieving you of that burden thank you very much
you
Feedback
hidden