In conversation | Geoffrey Hinton and Joel Hellermark

have

you reflected a lot on how to select

Talent or has that mostly been like

intuitive to you Ilia just shows up and

you're like this is a clever guy let's

let's work together or have you thought

a lot about that can we are we recording

should we should we roll This yeah let's

roll this okay we're good yeah

yeah

okay s is working

so I remember when I first got to K

melon from England in England at a

Research Unit it would get to be 6:00

and you'd all go for a drink in the pub

um at Caril melon I remember after I've

been there a few weeks it was Saturday

night I didn't have any friends yet and

I didn't know what to do so I decided

I'd go into the lab and do some

programming because I had a list machine

and you couldn't program it from home so

I went into the lab at about 9:00 on a

Saturday night and it was swarming all

the students were there and they were

all there because what they were working

on was the future they all believed that

what they did next was going to change

the course of computer science and it

was just so different from England and

so that was very refreshing take me back

to the very beginning Jeff at Cambridge

uh trying to understand the brain uh

what was that like it was very

disappointing so I did physiology and in

the summer term they were going to teach

us how the brain worked and it all they

taught us was how neurons conduct action

potentials which is very interesting but

it doesn't tell you how the brain works

so that was extremely disappointing I

switched to philosophy then I thought

maybe they'd tell us how the mind worked

um that was very disappointing I

eventually ended up going to Edinburgh

to do Ai and that was more interesting

at least you could simulate things so

you could test out theories and did you

remember what intrigued you about AI was

it a paper was it any particular person

that exposed you to those ideas I guess

it was a book I read by Donald Hebb that

influenced me a lot um he was very

interested in how you learn the

connection strengths in neural Nets I

also read a book by John Fon noyman

early on um who was very interested in

how the brain computes and how it's

different from normal computers and did

you get that conviction that this ideas

would work out at at that point or what

would was your intuition back at the

Edinburgh days it seemed to me there has

to be a way that the brain

learns and it's clearly not by having

all sorts of things programmed into it

and then using logical rules of

inference that just seemed to me crazy

from the outset um so we had to figure

out how the brain learned to modify

Connections in a neural net so that it

could do complicated things and Fon

Norman believed that churing believed

that so Forman and churing were both

pretty good at logic but they didn't

believe in this logical approach and

what was your split between studying the

ideas from from

neuroscience and just doing what seemed

to be good algorithms for for AI how

much inspiration did you take early on

so I never did that much study of

Neuroscience I was always inspired by

what I'd learned about how the brain

works that there's a bunch of neurons

they perform relatively simple

operations they're nonlinear um but they

collect inputs they wait them and then

they an output that depends on that

weighted input and the question is how

do you change those weights to make the

whole thing do something good it seems

like a fairly simple question what

collaborations do you remember from from

that time the main collaboration I had

at Carnegie melon was with someone who

wasn't at carnegy melon I was

interacting a lot with Terry sinowski

who was in Baltimore at John's Hopkins

and about once a month either he would

drive to Pittsburg or I drive to

Baltimore it's 250 miles away and we

would spend a weekend together working

on boltimore machines that was a

wonderful collaboration we were both

convinced it was how the brain worked

that was the most exciting research I've

ever done and a lot of technical results

came out that were very interesting but

I think it's not how the brain works um

I also had a very good collaboration

with um Peter Brown who was a very good

statistician and he worked on speech

recognition at IBM and then he came as a

more mature student to kind melon just

to get a PhD um but he already knew a

lot he taught me a lot about spee

and he in fact taught me about hidden

Markov models I think I learn more from

him than he learned from me that's the

kind of student you want and when he Tau

me about hidden Markov models I was

doing back propop with hidden layers

only they weren't called hidden layers

then and I decided that name they use in

Hidden Markov models is a great name for

variables that you don't know what

they're up to um and so that's where the

name hidden in neur NS came from me and

P decided that was a great name for the

hidden hidden L and your all Nets um but

I learned a lot from Peter about speech

take us back to when Ilia showed up at

your at your office I was in my office I

probably on a Sunday um and I was

programming I think and there was a

knock on the door not just any knock but

it won't

cutter it's sort of an urgent knock so I

went and answer to the door and this was

this young student there and he said he

was cooking Fries over the summer but

he'd rather be working in my lab and so

I said well why don't you make an

appointment and we'll talk and so Ilia

said how about now and that sort of was

Ila's character so we talked for a bit

and I gave him a paper to read which was

the nature paper on back

propagation and we made another meeting

for a week later and he came back and he

said I didn't understand it and I was

very disappointed I thought he seemed

like a bright guy but it's only the

chain rule it's not that hard to

understand and he said oh no no I

understood that I just don't understand

why you don't give the gradient to a

sensal a sensible function

Optimizer which took us quite a few

years to think about um and it kept on

like that with a he had very good his

raw intuitions about things were always

very good what do you think had enabled

those uh those intuitions for for Ilia I

don't know I think he always thought for

himself he was always interested in AI

from a young age um he's obviously good

at math so but it's very hard to know

and what was that collaboration between

between the two of you like what part

would you play and what part would Ilia

play it was a lot of fun um I remember

one occasion when we were trying to do a

complicated thing with producing maps of

data where I had a kind of mixture model

so you could take the same bunch of

similarities and make two maps so that

in one map Bank could be close to Greed

and in another map Bank could be close

to River um cuz in one map you can't

have it close to both right cuz River

and greed along wayon so we'd have a

mixture maps and we were doing it in mat

lab and this involved a lot of

reorganization of the code to do the

right Matrix multiplies and only got fed

up with that so he came one day and said

um I'm going to write a an interface for

Matlab so I program in this different

language and then I have something that

just converts it into Matlab and I said

no Ilia um that'll take you a month to

do we've got to get on with this project

don't get diverted by that and I said

it's okay I did it this

morning and that's that's quite quite

incredible and throughout those those

years the biggest shift wasn't

necessarily just the the algorithms but

but also the the skill how did you sort

of view that skill uh over over the

years Ilia got that intuition very early

so Ilia was always preaching that um you

just make it bigger and it'll work

better and I always thought that was a

bit of a copout do you going to have to

have new ideas too it turns out I was

basically right new ideas help things

like Transformers helped a lot but it

was really the scale of the data and the

scale of the computation and back then

we had no idea computers would get like

a billion times faster we thought maybe

they' get a 100 times faster we were

trying to do things by coming up with

clever ideas that would have just solved

themselves if we had had bigger scale of

the data and computation in about

2011 Ilia and another graduate student

called James Martins and

had a paper using character level

prediction so we took Wikipedia and we

tried to predict the next HTML character

and that worked remarkably well and we

were always amazed at how well it worked

and that was using a fancy Optimizer on

gpus and we could never quite believe

that it understood anything but it

looked as though it

understood and that just seemed

incredible can you take us through how

are do models trained to predict the

next word and why is it the wrong way of

of thinking about them okay I don't

actually believe it is the wrong way so

in fact I think I made the first

neuronet language model that used

embeddings and back propagation so it's

very simple data just

triples and it was turning each symbol

into an embedding then having the

embeddings interact to predict the

embedding of the next symbol and from

that predic the next symbol and then it

was back propagating through that whole

process to learn these triples and I

showed it could generalize um about 10

years later Yoshua Benji used a very

similar Network and showed it work with

real text and about 10 years after that

linguist started believing in embeddings

it was a slow process the reason I think

it's not just predicting the next symbol

is if you ask well what does it take to

predict the next symbol particularly if

you ask me a question and then the first

word of the answer is the next symbol um

you have to understand the question so I

think by predicting the next

symbol it's very unlike oldfashioned

autocomplete oldfashioned autocomplete

you'd store sort of triples of words and

then if you sort a pair of words you see

how often different words came third and

that way you can predict the next symbol

and that's what most people think auto

complete is like it's no longer at all

like that um to predict the next symbol

you have to understand what's been said

so I think you're forcing it to

understand by making it predict the next

symbol and I think it's understanding in

much the same way we are so a lot of

people will tell you these things aren't

like us um they're just predicting the

next symbol they're not reasoning like

us but actually in order to predict the

next symbol it's have going to have to

do some reasoning and we've seen now

that if you make big ones without

putting in any special stuff to do

reasoning they can already do some

reasoning and I think as you make them

bigger they're going to be able to do

more and more reasoning do you think I'm

doing anything else than predicting the

next symbol right now I think that's how

you're learning I think you're

predicting the next video frame um

you're predicting the next sound um but

I think that's a pretty plausible theory

of how the brain's learning what enables

these models to learn such a wide

variety of of fields what these big

language models are doing is they

looking for common structure and by

finding common structure they can encode

things using the common structure and

that more efficient so let me give you

an example if you ask

gp4 why is a compost heap like an atom

bomb most people can't answer that most

people haven't thought they think atom

bombs and compost heeps are very

different things but gp4 will tell you

well the energy scales are very

different and the time scales are very

different but the thing that's the same

is that when the compost Heep gets

hotter it generates heat faster and when

the atom bomb produces more NE neutrons

it produces more neutrons faster

and so it gets the idea of a chain

reaction and I believe it's understood

they're both forms of chain reaction

it's using that understanding to

compress all that information into its

weights and if it's doing that then it's

going to be doing that for hundreds of

things where we haven't seen the

analogies yet but it has and that's

where you get creativity from from

seeing these analogies between

apparently very different things and so

I think gp4 is going to end up when it

gets bigger being very creative I think

this idea that it's just just

regurgitating what it's learned just

pasing together text it's learned

already that's completely wrong it's

going to be even more creative than

people I think you'd argue that it won't

just repeat the human knowledge we've

developed so far but could also progress

beyond that I think that's something we

haven't quite seen yet we've started

seeing some examples of it but to a to a

large extent we're sort of still at the

current level of of of science what do

you think will enable it to go beyond

that well we've seen that in more

limited context like if you take Alpha

go in that famous competition with Leo

um there was move 37 where Alpha go made

a move that all the experts said must

have been a mistake but actually later

they realized it was a brilliant move um

so that was created within that limited

domain um I think we'll see a lot more

of that as these things get bigger the

difference with alphao as well was that

it was using reinforcement learning that

that subsequently sort of enabled it to

to go beyond the current state so it

started with imitation learning watching

how humans play the game and then it

would through selfplay develop Way

Beyond that do you think that's the

missing component of the I think that

may well be a missing component yes that

the the self-play in Alpha in Alpha go

and Alpha zero are are a large part of

why it could make these creative moves

but I don't think it's entirely

necessary

so there's a little experiment I did a

long time ago where you your training in

neuronet to recognize handwritten digits

I love that example the mest example and

you give it training data where half the

answers are

wrong um and the question is how well

will it

learn and you make half the answers

wrong once and keep them like that so it

can't average away the wrongness by just

seeing the same example but with the

right answer sometimes and the wrong

answer sometimes when it sees that

example half half of the examples when

it sees the example the answer is always

wrong and so the training data has 50%

error but if you train up back

propagation it gets down to 5% error or

less other words from badly labeled data

it can get much better results it can

see that the training data is wrong and

that's how smart students can be smarter

than their advisor and their advisor

tells them all this stuff

and for half of what their advisor tells

them they think no rubbish and they

listen to the other half and then they

end up smarter than the advisor so these

big neural Nets can actually do they can

do much better than their training data

and most people don't realize that so

how how do you expect this models to add

reasoning in into them so I mean one

approach is you add sort of the

heuristics on on top of them which a lot

of the research is doing now where you

have sort of Shan of thought you just

feedback it's reasoning um in into

itself and another way would be in the

model itself as you scale scale scale it

up what's your intuition around that so

my intuition is that as we scale up

these models I get better at reasoning

and if you ask how people work roughly

speaking we have these

intuitions and we can do reasoning and

we use the reasoning to correct our

intuitions of course we use the

intuitions during the reasoning to do

the reasoning but if the conclusion of

the reasoning conflicts with our in

itions we realize the intuitions need to

be changed that's much like in Alpha go

or Alpha zero where you have an

evaluation function um that just looks

at a board and says how good is that for

me but then you do the Monte Cara roll

out and now you get a more accurate idea

and you can revise your evaluation

function so you can train it by getting

it to agree with the results of

reasoning and I think these large

language models have to start doing that

they have to start training their Raw

intuitions about what should come next

by doing reasoning and realizing that's

not right and so that way they can get

more training data than just mimicking

what people did and that's exactly why

alphao could do this creative move 37 it

had much more training data because it

was using reasoning to check out what

the right next move should have been and

what do you think about multimodality so

we spoke about these analogies and often

the analogies are Way Beyond what we

could see it's discovering analogy that

are far beyond humans and at maybe

abstraction levels that we'll never be

able to to to understand now when we

introduce images to that and and video

and sound how do you think that will

change the models and uh how do you

think it will change the analogies that

it will be able to make um I think it'll

change it a lot I think it'll make it

much better at understanding spatial

things for example from language alone

it's quite hard to understand some

spatial things although remarkably gp4

can do that even before it was

multimodal um but when you make it

multimodal if you have it both doing

vision and reaching out and grabbing

things it'll understand object much

better if it can pick them up and turn

them over and so on so although you can

learn an awful lot from language it's

easier to learn if you multimodal and in

fact you then need less language and

there's an awful lot of YouTube video

for predicting the next frame so or

something like that so I think these

multimodule models are clearly going to

take over um you can get more data that

way they need less language so there's

really a philosophical point that you

could learn a very good model from

language alone but it's much easier to

learn it from a multimodal system and

how do you think it will impact the

model's reasoning I think it'll make it

much better at reasoning about space for

example reasoning about what happens if

you pick objects up if you actually try

picking objects up you're going to get

all sorts of training data that's going

to help do you think the human brain

evolved to work well with with language

or do you think language evolved to work

well with the human brain I think the

question of whether language evolved to

work with the brain or the brain evolved

to work with language I think that's a

very good question I think both happened

I used to think we would do a lot of

cognition without needing language at

all um now I've changed my mind a bit so

let me give you three different views of

language um and how it relates to

cognition there's the oldfashioned

symbolic view which is cognition

consists of having strings of symbols in

some kind of cleaned up logical language

where there's no ambiguity and applying

rules of inference and that's what

cognition is it's just these symbolic

manipulations on things that are like

strings of language symbols um so that's

one extreme view an opposite extreme

view is no no once you get inside the

head it's all vectors so symbols come in

you convert those symbols into big

vectors and all the stuff inside's done

with big vectors and then if you want to

produce output you produce symbols again

so there was a point in machine

translation in about

2014 when people were using neural

recurrent neural Nets and words will

keep coming in and that have a hidden

State and they keep accumulating

information in this hidden state so when

they got to the end of a sentence that

have a big hidden Vector that captures

the meaning of that sentence that could

then be used for producing the sentence

in another language that was called a

thought vector and that's a sort of

second view of language you convert the

language into a big Vector that's

nothing like language and that's what

cognition is all about but then there's

a third view which is what I believe now

which is that you take these

symbols and you convert the symbols into

embeddings and you use multiple layers

of that so you get these very rich

embeddings but the embeddings are still

to the symbols in the sense that you've

got a big Vector for this symbol and a

big Vector for that symbol and these

vectors interact to produce the vector

for the symbol for the next word and

that's what understanding is

understanding is knowing how to convert

the symbols into these vectors and

knowing how the elements of the vector

should interact to predict the vector

for the next symbol that's what

understanding is both in these big

language models and in our

brains and that's an example which is

sort of in between you're staying with

the symbols but you're interpreting them

as these big vectors and that's where

all the work is and all the knowledge is

in what vectors you use and how the

elements of those vectors interact not

in symbolic

rules um but it's not saying that you

get away from the symbols all together

it's saying you turn the symbols into

big vectors but you stay with that

surface structure of the symbols and

that's how these models are working and

that's I seem to be a more plausible

model of human thought too you were one

of the first folks to get idea of using

gpus and I know yansen loves you for

that uh back in 2009 you mentioned that

you told yansen that this could be a

quite good idea um for for training

training neural Nets take us back to

that early intuition of of using gpus

for for training neural Nets so actually

I think in about

2006 I had a former graduate student

called Rick zisy who's a very good

computer vision guy and I talked to him

and a meeting and he said you know you

ought to think about using Graphics

processing cards because they're very

good at Matrix multiplies and what

you're doing is basically all matric

multiplies so I thought about that for a

bit and then we learned about these

Tesla systems that had um four gpus in

and initially we just got um gaming gpus

and discovered they made things go 30

times faster and then we bought one of

these Tesla systems with 4 gpus and we

did speech on that and it worked very

well then in 2009 I gave a talk at nips

and I told a thousand machine learning

researches you should all go and buy

Nvidia gpus they're the future you need

them for doing machine learning and I

actually um then sent mail to Nvidia

saying I told a thousand machine

learning researchers to buy your boards

could you give me a free one and they

said no actually they didn't say no they

just didn't reply um but when I told

Jensen this story later on he gave me a

free

one that's uh that's very very good I I

think what's interesting is um as well

is sort of how gpus has evolved

alongside the the field so where where

do you think we we should go go next in

in the in the compute so my last couple

of years at Google I was thinking about

ways of trying to make analog

computation so that instead of using

like a megawatt we could use like 30

Watts like the brain and we could run

these big language models in analog

hardware and I never made it

work and but I started really

appreciating digital computation so if

you're going to use that low power

analog

computation every piece of Hardware is

going to be a bit different and the idea

is the learning is going to make use of

the specific properties of that hardware

and that's what happens with people all

our brains are different um so we can't

then take the weights in your brain and

put them in my brain the hardware is

different the precise properties of the