Geoffrey Hinton | On working with Ilya, choosing problems, and the power of intuition

Summary

The script features an interview with a renowned scientist discussing his journey from England to Carnegie Mellon, his early disappointment with traditional studies of the brain, and his shift to artificial intelligence (AI). He reflects on his experiences and collaborations in AI research, highlighting the development of neural networks and backpropagation. The discussion covers the evolution of AI, the importance of large models and multimodal data, and the potential societal impacts of AI advancements. The scientist emphasizes the importance of intuition, selecting talented students, and the future direction of AI research, including potential risks and ethical considerations.

Takeaways

• 🧠 The importance of environment: Moving from England to Carnegie Mellon, the speaker noticed a stark difference in work culture and environment, with students working late into the night believing their work would shape the future of computer science.
• 🔬 Disappointment in initial studies: The speaker found the study of the brain at Cambridge disappointing because it only covered basic neuron functions, leading him to switch to philosophy and eventually AI.
• 📘 Influential readings: Books by Donald Hebb and John von Neumann were pivotal in shaping the speaker's understanding and interest in how the brain learns and computes.
• 🤖 Collaborations: Key collaborations with Terry Sejnowski and Peter Brown were significant in the speaker's research, providing valuable insights and advancements in neural networks and speech recognition.
• 👨‍🎓 Notable students: Ilia, a standout student, impressed the speaker with his intuitive understanding of AI and neural networks, leading to impactful collaborations.
• 💡 Scale and innovation: Ilia's belief in the importance of scaling models was initially met with skepticism but proved crucial as larger models demonstrated significant improvements.
• 🧩 Understanding through prediction: The speaker believes that predicting the next symbol in language models forces a deeper understanding of context and reasoning.
• 🌐 Multimodal learning: Integrating multiple forms of data (text, images, video) into AI models will enhance their understanding and reasoning capabilities.
• ⚡ GPU revolution: The transition to using GPUs for training neural networks significantly accelerated AI research and development.
• 🔄 Digital immortality: Digital systems can share weights and knowledge efficiently, unlike human brains, leading to superior learning and knowledge dissemination.

Q & A

What was the speaker's initial impression of the academic environment at Carnegie Mellon compared to England?

The speaker found the environment at Carnegie Mellon to be very different and refreshing compared to England. Students at Carnegie Mellon were working late into the night because they believed their work was shaping the future of computer science, which was a stark contrast to the pub-going culture after 6:00 PM in England.

Why was the speaker disappointed with his initial studies in physiology and philosophy?

The speaker was disappointed because his studies in physiology only taught him about how neurons conduct action potentials, which didn't explain how the brain works as a whole. Similarly, philosophy didn't provide insights into how the mind worked, which was his ultimate interest.

What inspired the speaker to pursue AI research?

The speaker was inspired to pursue AI research after reading books by Donald Hebb and John von Neumann. Hebb's interest in learning connection strengths in neural nets and von Neumann's interest in brain computation intrigued the speaker and led him to Edinburgh to study AI.

How did the speaker's collaboration with Terry Sejnowski come about?

The speaker's collaboration with Terry Sejnowski began when they interacted frequently despite the distance between Pittsburgh and Baltimore. They would take turns visiting each other's city about once a month to work on Boltzmann machines, sharing a conviction that this was how the brain worked.

What was the significance of the speaker's collaboration with Peter Brown?

The collaboration with Peter Brown was significant because Peter, a statistician, taught the speaker about speech recognition and hidden Markov models. This collaboration was fruitful, with the speaker feeling that he learned more from Peter than Peter did from him.

How did Ilya Sutskever's initial interaction with the speaker influence their future collaboration?

Ilya Sutskever's initial interaction with the speaker demonstrated his eagerness and intuition for AI. Despite not understanding the paper on backpropagation initially, Ilya's question about why the gradient wasn't given to a sensible function optimizer showed his deep thinking, which led to a productive collaboration.

What was the speaker's view on the importance of scale in AI models?

The speaker believed that while new ideas like Transformers helped, the real shift in AI performance was due to the scale of data and computation. He mentioned that they didn't anticipate computers becoming a billion times faster, and with larger scale, models could achieve more without needing as many new ideas.

How does the speaker perceive the process of predicting the next word in language models?

The speaker believes that predicting the next word in language models is not just a mechanical process. It requires understanding the context, similar to how humans comprehend and generate language, which involves reasoning.

What role does the speaker see for multimodality in the future of AI models?

The speaker sees multimodality as a significant advancement for AI models. By incorporating images, video, and sound, models will improve in understanding spatial relationships and concepts that are difficult to grasp from language alone.

What was the speaker's intuition about the use of GPUs for training neural networks?

The speaker's intuition about using GPUs for training neural networks was based on their efficiency in performing matrix multiplications, which are fundamental to neural network computations. This led to significant speed improvements in training times.

How does the speaker view the relationship between language and cognition?

The speaker views language as a tool for cognition, where symbols are converted into embeddings that interact to predict subsequent symbols. This process of converting and interacting symbols is seen as central to both understanding and generating language.

What is the speaker's perspective on the potential of AI in healthcare?

The speaker sees AI in healthcare as a promising application, with the potential to significantly increase the availability and quality of medical care. AI could assist or replace doctors, leading to a situation where everyone could have personalized medical attention.

What is the speaker's approach to selecting research problems?

The speaker selects research problems based on intuition and a sense that a widely accepted idea might be wrong. He looks for opportunities to challenge conventional wisdom with simple demonstrations that can show why the prevailing view may not be accurate.

What does the speaker consider as the most promising direction in AI research today?

The speaker believes that training large models on multimodal data is a very promising direction. Even if the models are initially used for simple tasks like predicting the next word, the approach has great potential for future development.

What is the speaker's view on the importance of learning algorithms in achieving human-level intelligence?

The speaker believes that while backpropagation is a fundamentally correct and successful approach for learning, there may be alternative learning algorithms that could also achieve human-level intelligence. However, he acknowledges that backpropagation has proven to be highly effective.

What achievement from the speaker's career is he most proud of?

The speaker is most proud of the development of the learning algorithm for Boltzmann machines. He considers it elegant, even if it may not be practical, and it was a project he greatly enjoyed working on.

Outlines

00:00

🌟 Early Inspirations and AI Explorations

The speaker reminisces about their early experiences at Carnegie Mellon and the inspiring work ethic of students there. They express disappointment with their initial studies in physiology and philosophy, which led them to pursue AI at Edinburgh. The influence of Donald Hebb and John von Neumann on their interest in neural networks and brain function is highlighted. The speaker also shares their early conviction that the brain learns without pre-programmed logical rules, a belief that motivated their research in AI.

05:04

🤝 Collaborations and Intuitions in AI Development

This paragraph delves into the speaker's collaborations, particularly with Terry Sejnowski and Peter Brown, and the insights gained from them. The speaker reflects on the importance of collaborations in developing their understanding of AI, the adoption of the term 'hidden layers' from hidden Markov models, and the evolution of their ideas on how the brain might work. The paragraph also touches on the speaker's initial disappointment with a student, Ilia, who later proved to have profound intuitions about AI, challenging conventional approaches to optimization.

10:06

🚀 The Advent of Backpropagation and Scaling in AI

The speaker recounts the early days of backpropagation and their work on neural networks. They discuss the evolution of AI from clever ideas to the realization that scale in data and computation is crucial. The introduction of character-level prediction using large datasets like Wikipedia is highlighted, along with the surprising effectiveness of such models. The speaker also addresses the misconception that AI models are merely predicting the next word, arguing that understanding and reasoning are inherent in the process.

15:07

🧠 The Intersection of Neuroscience and AI

In this section, the speaker explores the relationship between neuroscience and AI, discussing the brain's learning processes and the inspiration drawn from it for developing AI algorithms. They emphasize the importance of embeddings and backpropagation in creating language models that can generalize and understand context. The speaker also speculates on the potential for AI to develop creativity and reasoning abilities beyond human levels as models grow larger and more complex.

20:08

🔄 The Role of Reinforcement Learning and Analogies in AI

The speaker discusses the impact of reinforcement learning, as exemplified by AlphaGo's move 37, and the potential for AI to develop creative solutions within limited domains. They also touch on the concept of 'fast weights' and the brain's ability to adapt quickly, suggesting that AI could benefit from incorporating similar mechanisms. The speaker further elaborates on the potential of multimodal models to enhance understanding and reasoning in AI.

25:10

🌐 The Future of AI and Its Impact on Society

In this paragraph, the speaker contemplates the future applications of AI, particularly in healthcare and materials engineering, while also expressing concern about the potential misuse of AI by bad actors. They acknowledge the balance between the positive and negative impacts of AI and the importance of international cooperation in advancing the field responsibly.

30:11

🤖 The Evolution of AI and Lessons from Neuroscience

The speaker reflects on the evolution of AI and the lessons learned from neuroscience, emphasizing the importance of multiple timescales in learning and the potential for AI to incorporate 'fast weights' similar to the brain. They also discuss the impact of their understanding of the brain on AI development and the philosophical implications of AI's ability to reason and understand.

35:11

💡 The Intuitive Process of Selecting Talent in AI Research

In this section, the speaker shares insights into their intuitive process of selecting talent for AI research, highlighting the importance of recognizing deep insights and creativity in potential collaborators. They discuss the value of having a variety of student types in a lab and trust their gut intuition when identifying promising individuals.

40:11

🛠 The Importance of Intuition and Diverse Approaches in AI

The speaker emphasizes the role of intuition in developing good ideas and the importance of not accepting everything they are told. They advocate for having a strong framework for understanding reality and being able to reject information that doesn't fit. The speaker also discusses the potential for a variety of learning algorithms to achieve human-level intelligence.

45:12

🏆 Reflecting on Achievements and the Journey of AI Research

In the concluding paragraph, the speaker reflects on their proudest achievement—the development of the learning algorithm for Boltzmann machines—and their current musings, including the question of whether the brain uses backpropagation. They express a lifelong curiosity about the brain's learning capabilities and acknowledge the unexpected positive outcomes of their research journey.

Keywords

💡Intuition

Intuition refers to the ability to understand or know something immediately, without the need for conscious reasoning. In the context of the video, it is highlighted as a key trait in selecting talent, where the speaker mentions that some individuals, like Ilia, have a strong intuitive grasp of complex concepts. The video suggests that intuition can play a significant role in the field of AI, as it often guides researchers towards innovative ideas and solutions.

💡Backpropagation

Backpropagation is a method used in artificial neural networks to calculate the gradient of the loss function with respect to the weights. It is central to the training process of many AI models. The script discusses the significance of backpropagation in understanding how neural networks learn and the speculation about its biological counterpart in the human brain, suggesting that it might be a fundamental process for learning in both AI and neural systems.

💡Neural Networks

Neural networks are a set of algorithms designed to recognize patterns. They are inspired by the human brain and are composed of interconnected nodes or 'neurons'. The video script discusses the evolution of neural networks, from their initial disappointing results in simulating the brain to their current state as powerful tools in AI, capable of complex tasks like language processing and pattern recognition.

💡AI

AI, or artificial intelligence, is the field of study focused on creating machines that can perform tasks that would typically require human intelligence, such as problem-solving, learning, and understanding language. The video script frequently references AI, discussing its development, the influence of individuals in the field, and the philosophical implications of creating intelligent machines.

💡Embeddings

In the context of AI, embeddings are vector representations of words or symbols in a reduced dimensional space. They capture semantic meaning and are used in various NLP tasks. The script mentions embeddings as a crucial part of language models, allowing AI to understand and predict text based on the context provided by these vector representations.

💡Multimodal Data

Multimodal data refers to information that is derived from multiple different types of input, such as text, images, and sound. The script discusses the potential of training AI models on multimodal data, suggesting that this approach could lead to more robust and versatile AI systems capable of understanding and predicting a wider range of human experiences and interactions.

💡Boltzmann Machines

Boltzmann machines are a class of stochastic artificial neural networks capable of learning from data. The script mentions the speaker's past work on Boltzmann machines, highlighting the elegance of their learning algorithm, even though it may not be how the brain actually learns or the most practical approach in AI.

💡Cognitive Dissonance

Cognitive dissonance is the mental discomfort experienced by a person who holds two or more contradictory beliefs. In the script, the concept is alluded to when discussing the process of selecting talent and developing intuition, suggesting that individuals who can reconcile conflicting information effectively may have better intuition.

💡Hidden Markov Models

Hidden Markov Models (HMMs) are statistical models that represent a set of possible output sequences, each with an associated statistical model, e.g., a sequence of words in speech recognition. The script mentions HMMs as an example of a technical concept that the speaker learned from a student, demonstrating the collaborative and educational nature of scientific research.

💡Reinforcement Learning

Reinforcement learning is a type of machine learning where an agent learns to make decisions by performing actions in an environment to achieve a goal. The script refers to AlphaGo, a program that uses reinforcement learning to play the game of Go, as an example of how AI can surpass human levels of play through self-play and learning from its mistakes.

💡Analogies

Analogies are comparisons between two things that are different in most respects but alike in some particular way. The script discusses the ability of AI models to make analogies as a sign of advanced understanding and creativity, suggesting that this capacity could lead to AI systems that are capable of original thought and innovation.

Highlights

The importance of selecting talent based on intuition and capability, as demonstrated by the story of Ilya's recruitment.

The cultural difference in work ethic and motivation between England and Carnegie Mellon, highlighting the dedication to future advancements in computer science.

Disappointment with traditional education in physiology and philosophy, leading to a switch to AI and the excitement of simulating and testing theories.

Influence of Donald Hebb's work on learning in neural networks and the early interest in how the brain computes differently from traditional computers.

The early conviction that the brain learns without pre-programmed logical rules, and the pursuit to understand neural net modifications for complex tasks.

The collaboration with Terry Sejnowski and the excitement of working on what was believed to be the future of understanding brain function.

Learning from Peter Brown about speech recognition and the origin of the term 'hidden layers' in neural networks.

Ilia's unique intuition for AI and the significant discussion about the use of gradient in optimization, leading to years of contemplation.

The fun and productive collaboration between the speaker and Ilia, including the story of writing an interface for Matlab in a morning.

The evolution of understanding in AI, from the belief in the necessity of new ideas to the realization of the importance of scale in data and computation.

The development and success of character-level prediction using Wikipedia, showcasing the capabilities of neural networks.

The debate on whether AI models are simply predicting the next word or actually understanding and reasoning like humans.

The potential for large language models to discover analogies and be creative beyond human capabilities.

The impact of multimodality on AI models, suggesting improvements in understanding spatial relationships and reasoning.

The philosophical implications of language in cognition and the evolution of views on how symbols are processed in the brain.

The early adoption and advocacy for using GPUs in training neural networks, highlighting the significant increase in computational efficiency.

The exploration of analog computation and the potential for AI to mimic the brain's efficiency in power usage.

The potential for AI in healthcare, emphasizing the vast need and opportunity for improvement in medical services.

Concerns about the misuse of AI by bad actors for harmful purposes, such as killer robots, public opinion manipulation, and mass surveillance.

Reflections on the process of selecting talent in AI research and the importance of intuition and technical strength.

Transcripts

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

individual ual neurons are different the

learning used to make has learned to

make use of all that and so we're mortal

in the sense that the weights in my

brain are no good for any other brain

when I die those weights are useless um

we can get information from one to

another rather

inefficiently by I produce sentences and

you figure out how to change your weight

so you would have said the same thing

that's called distillation but that's a

very inefficient way of communicating

knowledge and with digital systems

they're immortal because once you got

some weights you can throw away the

computer just store the weights on a

tape somewhere and now build another

computer put those same weights in and

if it's digital it can compute exactly

the same thing as the other system did

so digital systems can share weights and

that's incredibly much more efficient if

you've got a whole bunch of digital

systems and they each go and do a tiny

bit of

learning and they start with the same

weights they do a tiny bit of learning

and then they share their weights again

um they all know what all the others

learned we can't do that and so they're

far superior to us in being able to

share knowledge a lot of the ideas that

have been deployed in the field are very

old school ideas uh it's the ideas that

have been around the Neuroscience for

forever what do you think is sort of

left to to to apply to the systems that

we develop so one big thing that we

still have to catch up with Neuroscience

on is the time scales for changes so in

nearly all the neural Nets there's a

fast time scale for changing activities

so input comes in the activities the

embedding vectors all change and then

there's a slow time scale which is

changing the weights and that's

long-term learning and you just have

those two time scales in the brain

there's many time scales at which

weights change so for example if I say

an unexpected word like cucumber and now

5 minutes later you put headphones on

there's a lot of noise and there's very

faint words you'll be much better at

recognizing the word cucumber because I

said it 5 minutes ago so where is that

knowledge in the brain and that

knowledge is obviously in temporary

changes to synapsis it's not neurons are

going cucumber cucumber cucumber you

don't have enough neurons for that it's

in temporary changes to the weights and

you can do a lot of things with

temporary weight changes fast what I

call fast weights we don't do that in

these neural models and the reason we

don't do it is because if you have

temporary changes to the weights that

depend on the input data then you can't

process a whole bunch of different cases

at the same time at present we take a

whole bunch of different strings we

stack them stack them together and we

process them all in parallel because

then we can do Matrix Matrix multiplies

which is much more efficient and just

that efficiency is stopping us using

fast weights but the brain clearly uses

fast weights for temporary memory and

there's all sorts of things you can do

that way that we don't do at present I

think that's one of the biggest things

we have to learn I was very hopeful that

things like graph core um if they went

sequential and did just online learning

then they could use fast weights

um but that hasn't worked out yet I

think it'll work out eventually when

people are using conductances for

weights how has knowing how this models

work and knowing how the brain works

impacted the way you you think I think

there's been one big impact which is at

a fairly abstract level which is that

for many

years people were very scornful about

the idea of having a big random neural

net and just giving a lot of training

data and it would learn to do

complicated things if you talk to

statisticians or linguists or most

people in AI they say that's just a pipe

dream there's no way you're going to

learn to really complicated things

without some kind of innate knowledge

without a lot of architectural

restrictions it turns out that's

completely wrong you can take a big

random neural network and you can learn

a whole bunch of stuff just from data um

so the idea that stochastic gradient

descent to adjust the repeatedly adjust

the weights using a gradient that will

learn things and we'll learn big

complicated things that's been validated

by these big models and that's a very

important thing to know about the brain

it doesn't have to have all this innate

structure now obviously it's got a lot

of innate structure but it certainly

doesn't need innate structure for things

that are easily

learned and so the sort of idea coming

from Chomsky that you won't you won't

learn anything complicated like language

unless it's all kind of wired in already

and just matures that idea is now

clearly nonsense I'm sure shumsky would

appreciate you calling his ideas

nonsense well I think actually I think a

lot of chs's political ideas are very

sensible and I'm was struck by how how

come someone with such sensible ideas

about the Middle East could be so wrong

about

Linguistics what do you think would make

these models simulate consciousness of

of humans more effectively but imagine

you had the AI assistant that you've

spoken to in your entire life and

instead of that being you know like chat

today that sort of deletes the memory of

the conversation and you start fresh all

of the time okay it had

self-reflection at some point you you

pass away and you tell that to to the

assistant do you think I me not me

somebody else tells that toist yeah you

would it would be difficult for you to

tell that to the assistant um do you

think that assistant would would feel at

that point yes I think they can have

feelings too so I think just as we have

this inner theater model for perception

we have an inthat model for feelings

they're things that I can experience but

other people can't um

I think that model is equally wrong so I

think suppose I say I feel like punching

Gary on the nose which I often do let's

try and Abstract that away from the idea

of an inner theater what I'm really

saying to you is um if it weren't for

the inhibition coming from my frontal

loes I would perform an action so when

we talk about feelings we really talking

about um actions we would perform if it

weren't for um con straints and that

really that's really what feelings are

the actions we would do if it weren't

for

constraints um so I think you can give

the same kind of explanation for

feelings and there's no reason why these

things can't have feelings in fact in

1973 I saw a robot having an emotion so

in Edinburgh they had a robot with two

grippers like this that could assemble a

toy car if you put the pieces separately

on a piece of green felt um but if you

put them in a pile his vision wasn't

good enough to figure out what was going

on so it put his grip whack and it

knocked them so they were scattered and

then it could put them together if you

saw that in a person you say it was

crossed with the situation because it

didn't understand it so it destroyed

it that's

profound you uh we spoke previously you

described sort of humans and and and and

the llms as analogy machines what do you

think has been the most powerful

analogies that you found throughout your

life oh in throughout my life um woo I

guess probably an a sort of weak analogy

that's influenced me a lot is um the

analogy between religious belief and

between belief in symbol

processing so when I was very young I

was confronted I came from an atheist

family and went to school and was

confronted with religious belief and it

just seemed nonsense to me it still

seems nonsense to me um and when I saw

symbol processing as an explanation how

people worked um I thought it was just

the same

nonsense I don't think it's quite so

much nonsense now because I think

actually we do do symbol processing it's

just we do it by giving these big

embedding vectors to the symbols but we

are actually symbol processing um but

not at all in the way people thought

where you match symbols and the only

thing is symbol has is it's identical to

another symbol or it's not identical

that's the only property a symbol has we

don't do that at all we use the context

to give embedding vectors to symbols and

then use the interactions between the

components of these embedding vectors to

do thinking but there's a very good

researcher at Google called Fernando

Pereira who said yes we do have symbolic

reasoning and the only symbolic we have

is natural language natural language is

a symbolic language and we reason with

it and I believe that now you've done

some of the most meaningful uh research

in the history of of computer science

can you walk us through like how do you

select the right problems to to work on

well first let me correct you me and my

students have done a lot of the most

meaningful things and it's mainly been a

very good collaboration with students

and my ability to select very good

students and that came from the fact

that were very few people doing neural

Nets in the 70s and 80s and 90s and

2000s and so the few people doing your

nets got to pick the very best students

so that was a piece of luck but my way

of selecting problems is

basically well you know when scientists

talk about how they work they have

theories about how they work which

probably don't have much to do with the

truth but my theory is that

I look for something where everybody's

agreed about something and it feels

wrong just there's a slight intuition

there's something wrong about it and

then I work on that and see if I can

elaborate why it is I think it's wrong

and maybe I can make a little demo with

a small computer program that shows that

it doesn't work the way you might expect

so let me take one example um most

people think that if you add noise to a

neural net is going to work worse um if

for example each time you put a training

example through

you make half of the neurons be silent

it'll work worse actually we know it'll

generalize better if you do that

and you can demonstrate that um in a

simple example that's what's nice about

computer simulation you can show you

know this idea you had that adding noise

is going to make it worse and sort of

dropping out half the neurons will make

it work worse which you will in the

short term but if you train it with like

that in the end it'll work better you

can demonstrate that with a small

computer program and then you can think

hard about why that is and how it stops

big elaborate co- adaptations um but

that I think that that's my method of

working find something that sounds

suspicious and work on it and see if you

can give a simple demonstration of why

it's wrong what sounds suspicious to you

now well that we don't use fast weight

sounds suspicious that we only have

these two time scales that's just wrong

that's not at all like the brain um and

in the long run I think we're going to

have to have many more time scans so

that's an example there and if you had

if you had your group of of students

today and they came to you and they said

so the Hamming question that we talked

about previously you know what's the

most important problem in in in your

field what would you suggest that they

take on and work on on next we spoke

about reasoning time scales what would

be sort of the highest priority Problem

that that you'd give them for me right

now it's the same question I've had for

the last like 30 years or so which is

does the brain do back propagation I

believe the brain is getting gradients

if you don't get gradients your learning

is just much worse than if you do get

gradients but how is the brain getting

gradients and is it

somehow implementing some approximate

version of back propagation or is it

some completely different technique

that's a big open question and if I kept

on doing research that's what I would be

doing research on and when you look back

at at your career now you've been right

about so many things but what were you

wrong about that you wish you sort of

spent less time pursuing a certain

direction okay those are two separate

questions one is what were you wrong

about and two do you wish you'd less

spent less time on it I think I was

wrong about Boltz machines and I'm glad

I spent a long time on it there are much

more beautiful theory of how you get

gradients than back propagation back

propagation is just ordinary and

sensible and it's just a chain rule B

machines is clever and it's a very

interesting way to get gradients and I

would love for that to be how the brain

works but I think it isn't did you spend

much time imagining what would happen

post the systems developing as as well

did you have an idea that okay if we

could make these systems work really

well we could you know democratize

education we could make knowledge way

more accessible um we could solve some

tough problems in in in medicine or was

it more to you about understanding the

Brin yes I I sort of feel scientists

ought to be doing things that are going

to help Society but actually that's not

how you do your best research you do

your best research when it's driven by

curiosity you just have to understand

something um much more recently I've

realized these things could do a lot of

harm as well as a lot of good and I've

become much more concerned about the

effects they're going to have on society

but that's not what was motivating me I

just wanted to understand how on Earth

can the brain learn to do things that's

what I want to know and I sort of failed

as a side effect of that failure we got

some nice engineering

but yeah it was a good good good failure

for the world if you take the lens of

the things that could go really right

what what do you think are the most

promising

applications I think Health Care is

clearly uh a big one um with Health Care

there's almost no end to how much Health

Care Society can absorb if you take

someone old they could use five doctors

fulltime um so when AI gets better than

people at doing things um you'd like it

to get better in areas where you could

do with a lot more of that stuff and we

could do with a lot more doctors if

everybody had three doctors of their own

that would be great and we're going to

get to that point um so that's one

reason why Healthcare is good there's

also just a new engineering developing

new materials for example for better

solar panels or for superc conductivity

or for just understanding how the Body

Works um there's going to be huge

impacts there those are all going to be

be good things what I worry about is Bad

actors using them for bad things we've

facilitated people like Putin or Z or

Trump

using AI for Killer Robots or for

manipulating public opinion or for Mass

surveillance and those are all very

worrying things are you ever concerned

that slowing down the field could also

slow down the positives oh absolutely

and I think there's not much chance that

the field will slow down partly because

it's International and if one country

slows down the other countries aren't

going to slow down so there's a race

clearly between China and the US and

neither is going to slow down so yeah I

don't I mean there was this partition

saying we should slow down for six

months I didn't sign it just because I

thought it was never going to happen I

maybe should have signed it because even

though it was never going to happen it

made a political point it's often good

to ask for things you know you can't get

just to make a point um but I didn't

think we're going to slow down and how

do you think that it will impact the AI

research process uh having uh this

assistance so I think it'll make it a

lot more efficient a research will get a

lot more efficient when you've got these

assistants that help you program um but

also help you think through things and

probably help you a lot with equations

too have you reflected much on the

process of selecting Talent has that

been mostly intuitive to you like when

Ilia shows up at the door you feel this

is smart guy let's work together so for

selecting Talent um sometimes you just

know so after talking to Ilia for not

very long he seemed very smart and then

talking him a bit more he clearly was

very smart and had very good intuitions

as well as being good at math so that

was a no-brainer there's another case

where I was at a NPS conference um we

had a poster and I someone came up and

he started asking questions about the

poster and every question he asked was a

sort of deep insight into what we'd done

wrong um and after 5 minutes I offered

him a postto position that guy was David

McKai who was just brilliant and it's

very sad he died but he was it was very

obvious you'd want him um other times

it's not so obvious and one thing I did

learn was that people are different

there's not just one type of good

student um so there's some students who

aren't that creative but are technically

extremely strong and will make anything

work there's other students who aren't

technically strong but are very creative

of course you want the ones who are both

but you don't always get that but I

think actually in the lab you need a

variety of different kinds of graduate

student but I still go with my gut

intuition that sometimes you talk to

somebody and they're just very very they

just get it and those are the ones you

want what do you think is the reason for

some folks having better intuition do

they just have better training data than

than others or how can you develop your

intuition I think it's partly they don't

stand for nonsense so here's a way to

get bad intuitions believe everything

you're told that's fatal you have to be

able to I think here's what some people

do they have a whole framework for

understanding reality and when someone

tells them something they try and sort

of figure out how that fits into their

framework and if it doesn't they just

reject it and that's a very good

strategy um people who try and

incorporate whatever they're told end up

with a framework that's sort of very

fuzzy and sort of can believe everything

and that's useless so I think actually

having a strong view of the world and

trying to manipulate incoming facts to

fit in with your view obviously it can

lead you into deep religious belief and

fatal flaws and so on like my belief in

boltzman machines um but I think that's

the way to go if you got good intuitions

you can trust you should trust them if

you got bad intuitions it doesn't matter

what you do so you might as well trust

them a very very good very good point

when when you look at the the types of

research that's that's that's being done

today do you think we're putting all of

our eggs in one basket and we should

diversify our ideas a bit more in in the

field or do you think this is the most

promising Direction so let's go all in

on it

I think having big models and training

them on multimodal data even if it's

only to predict the next word is such a

promising approach that we should go

pretty much all in on it obviously

there's lots and lots of people doing it

now and there's lots of people doing

apparently crazy things and that's good

um but I think it's fine for like most

of the people to be following this path

because it's working very well do you

think that the learning algorithms

matter that much or is it just a skill

are there basically millions of ways

that we could we could get to human

level in in intelligence or are there

sort of a select few that we need to

discover yes so this issue of whether

particular learning algorithms are very

important or whether there's a great

variety of learning algorithms that'll

do the job I don't know the answer it

seems to me though that back propagation

there's a sense in which it's the

correct thing to do getting the gradient

so that you change a parameter to make

it work better that seems like the right

thing to do and it's been amazingly

successful there may well be other

learning algorithms that are alternative

ways of getting that same gradient or

that are getting the gradient to

something else and that also work

um I think that's all open and a very

interesting issue now about whether

there's other things you can try and

maximize that will give you good systems

and maybe the brain's doing that because

it's

easier but backprop is in a sense the

right thing to do and we know that doing

it works really

well and one last question when when you

look back at your sort of Decades of

research what are you what are you most

proud of is it the students is it the

research what what makes you most proud

of when you look back at at your life's

work the learning algorithm for

boltimore machines so the learning

algorithm for Boltz machines is

beautifully elegant it's maybe hopeless

in practice um but it's the thing I

enjoyed most developing that with Terry

and it's what I'm proudest of um even if

it's

[Music]

wrong what questions do you spend most

of your time thinking about now is it

the um what what should I watch on

Netflix