📱

Get Our Mobile App

Take your business learning on the go!

Download on the App StoreGet it on Google Play

He Co-Invented the Transformer. Now: Continuous Thought Machines [Llion Jones / Luke Darlow]

Machine Learning Street Talk1:12:40

Transcription

despite the fact that I was involved in

inventing the Transformer luckily um no

one's been working on them as long as I

have rights with maybe the exception of

the other se seven authors. So I

actually made the decision uh earlier

this year that I'm going to drastically

reduce the amounts of of research that

I'm doing specifically on the

transformer because of the feeling that

I have that it's it's an oversaturated

space, right? It's not that there's no

more interesting things to be done with

them. And I'm going to make use of the

opportunity to do something different,

right? To actually turn up the amount of

exploration that I'm doing in my

research.

We just released the continuous thought

machine. It's a spotlight at Europe's

2025 this year. You should care about it

because it has native adaptive compute.

It's a new way of building a recurrent

model that uses [music] higher level

concepts for neurons and a

synchronization as a representation that

lets us solve problems in ways that seem

more human by being biologically and

nature inspired.

The atmosphere in AI research was

actually quite different back during the

Transformer uh years um because it

doesn't feel like something [music]

similar could actually happen right now.

because of the reduced amount of freedom

that we have, right? [music] The

Transformers was very very bottom up,

right? It's not that somebody had this

grand plan that came down from on high

that this is what we should be working

on. It was a bunch of people talking

over lunch, thinking about [music] what

the current problems are and how to

solve them and having the freedom to

have, you know, literally months to

dedicate to just trying this [music]

idea and having this this new

architecture fall out. [music]

We've spent hundreds of millions of

dollars. The biggest sort of evolution

based search is probably in the tens of

thousands. We have all this compute.

What happens? What happens if you scale

up these search algorithms? And I'm sure

you'll find something interesting, you

know, when someone eventually does bite

that bullet and really scale [music] up

these evolutionary sort of a life

experiments because I pitched it in an

environment where people were just going

all in on this one technology. I got

zero interest.

So now I have my own company and I can

pursue those directions.

This podcast is supported by Cyber Fund.

>> Hey folks, I'm Omar, product and design

lead at Google DeepMind. We just

launched a revamped vibe coding

experience in AI Studio that lets you

mix and match AI capabilities [music] to

turn your ideas into reality faster than

ever. Just describe your app and Gemini

will automatically wire up the right

models and APIs for you. And if you need

a spark, hit I'm feeling lucky and we'll

help you get started. Head to

a.studio/build studio/build to create

your first app.

>> Two for AI Labs is a research [music]

lab based in Zurich. They've got a team

of amazing ML engineers and research

scientists. They're doing some really

cool stuff. If you look at their

website, [music] for example, you can

see what their approach was for winning

the ARC AGI 3 pub uh competition which

closed out a few months ago and they are

[music] hiring amazing ML engineers and

research scientists. They also care

deeply about AI safety. So if any of

that is a fit for you, please go to Two

for [music] AOLABS and uh give it a go.

The audience will know I'm a huge fan of

Kenneth Stanley's ideas. So his book,

Why Greatness Cannot Be Planned, changed

my life. It was absolutely insane. And

what he was speaking to is that we need

to allow people to follow their own

gradient of interest unfettered by

objectives and committees and and so on.

Because that is how we do epistemic

foraging. that when you have too many

agendas involved in the mix, you kind of

end up with a gray goo and you don't

discover, you know, interesting novelty

and diversity. And I suppose that's

basically the thesis of of your company,

Sakana, is to lean into those ideas.

>> Yes, exactly. Um, at the company, we're

a massive fan of that book. We're we're

hoping to have him come and talk at our

company next week, actually. And um it's

a philosophy that we we do talk about

internally, right? We have copies of the

books in including the the recent

Japanese translation. As you know, one

of the co-founders, one of my my main

jobs, one of the main things that I have

to keep doing for this company is making

sure that we protect the freedom that

the researchers currently have, right?

Because it's it's it's a it's a

privilege really that we have the

resources to be able to do that. And

inevitably, as I've seen happen, as the

company grows, more and more pressure

comes in and it narrows the freedom. But

I think because, you know, we we believe

in this philosophy so strongly, I'm

hoping that we can give people all the

research freedom that we do now um for

as long as possible.

>> And what are those processes that

curtail freedom as a company matures? I

mean, how would you describe that? It's

great that there's never been so much

interest and people and talent and

resources and money in the industry,

but unfortunately that just increases

the amount of pressure people have in

order to compete with all the other

people working on it and trying to get

the the value out of this technology and

making money.

And I think that's what just happens,

right? As a startup, you have a a

feeling of, you know, excitement and

trying something new. And right at the

beginning, you have a bit of a runway.

So, you have the freedom to try

different things. But inevitably, people

are starting to ask for returns on their

investments or they're expecting you to

churn out some product. And this just

unfortunately

reduces the uh the the creativity that

that researchers have because you know

the the the the pressures to publish or

the pressure to to create technology

that's actually useful for the products

that we have goes up and so the feeling

of autonomy I think starts to go down.

But you know I literally tell people

when they start working for the company

I want you to work on what you think is

interesting and important and I mean it

there there is I mean in YouTube there's

a phenomenon called audience capture

>> right

>> and I think there might be a phenomenon

called technology capture which is that

in the early days of Google it was quite

open-ended and I mean transformers is

now the ubiquitous backbone of all AI

technology and it's a huge achievement

that that you're involved in But I mean

there's a similar story with with Open

AI. They're now starting to see all of

these commercialization opportunities.

They they can I mean they're going to

become LinkedIn. They're going to become

an application platform. They're going

to become a search platform. They're

going to become a social network. And

and I guess this could happen to you

guys that there's a very strong chance,

especially with your new paper that

we're going to talk about today, this

continuous thought machines. It it could

be a revolutionary technology, but then

it will become obvious how it could be

commercialized. And that's how those

pressures come in.

>> I I like the I like the audience capture

analogy.

I think um there's definitely been some

kind of capture by large language

models, right? They they worked so well

that everyone wanted to work on them.

And I'm really worried that we're kind

of stuck in this local minimum now,

right? and we sort of need to try to try

to escape it. So, we spoke about the

transformers, but there's a there's a

time just before the transformers that

I'd like to talk about because I think

it's quite illustrative. So, of course,

the the main technology before

transformers was recurrent neural

networks, right? And there was a similar

feeling, right? When recurrent neural

networks came in and we we you know, we

discovered this new sort of sequence of

sequence learning, that was also a

massive breakthrough, right? the the the

the the translation quality went up

massively, right? Um voice recognition

uh quality went up massively. And there

was this a similar sort of feeling then

of like okay yes we've you know we found

the technology and we just need to sort

of perfect this technology and back then

even my my favorite uh task was uh

character level language modeling right

so every time a new RNN based character

level language modeling paper came out I

got quite excited right um I you know

I'd want to like quickly read the paper

like okay how did they you know how did

they get the improvements

But the papers were always these just

these slight modifications on the same

architecture, right? It was LSTMs and

GRUs and maybe um initializing it with

the identity matrix to so that you could

use the relu function or like maybe if

you put the gate in a different place or

if you if you layer them in a slightly

different way or if you had gating going

upwards as well as as sideways. Um, and

I remember one of my favorites was this

uh like hierarchical LSTM where it would

actually decide to compute or not

compute the different layers. And if you

trained on Wikipedia and you looked at

the structure of when it was decided to

compute or not compute, it kind of

looked like the structure of of the the

sentences were actually being picked up

by the model. And I used to love that

sort of stuff, right? Um, but the the

the the improvements were always like

1.26 bits per character, 1.25 bits per

character, 1.24. That was a result that

was publishable, right? That was

exciting.

But then after the transformer the team

that I went on to afterwards right we

applied for the first time very deep

transformer models decoder only

transformer models to language modeling

and we immediately got something like

1.1

uh right so so something that was so

good that people actually come to our

desk and politely tell us like uh I

think you you made a error like a

calculation do you think it's nats not

bits per character and we're like no no

no we you know it really is the the the

correct the correct number. What struck

me later is that all of a sudden all of

that research and to be clear very good

research was suddenly made completely

redundant.

>> Yes.

>> Right. All of those endless permutations

to to RNN's

were suddenly seemingly a waste of time.

We're kind of in the situation right now

where a lot of the papers are just

taking the same architecture

and making these endless amount of

different tweaks of like you know where

to put them normalization layer and

slightly different ways of training them

and and we might be wasting the time in

exactly the same way right like I

personally don't think we're done right

I don't think that this is the final

architecture and we just need to keep

scaling up there's some breakthrough

that will occur at some point and then

it will once again become obvious that

we're kind of wasting a lot of time

right now.

>> Yeah. So we are a victim of our own

success and this basin of attraction

there are so many basins of attraction.

Sarah Hooker spoke about the hardware

lottery and this is a kind of

architecture lottery and it it it

actually made me think of the um

agricultural revolution which is that

this kind of phase change happened and

all of the folks that had these skills

that were so necessary, these diverse

skills for living and surviving, they

died out. And that's actually quite

paradoxical because we need those skills

to take the next step. M

>> and so we we're now in this regime we've

got the term foundation model and the

implication is that you can do anything

with a foundation model in the corporate

world we used to have data scientists

you know they were ML engineers doing

these architectural tweaks even in you

know midsize enterprise and now we just

have AI engineers who are just doing

prompt engineering and so on. So you're

saying that the fundamental skills that

we need to be diverse to think of new

solutions and new architectures, they're

dying out. I think I'm going to disagree

with that. I think the problem is we

have plenty of very talented

uh very creative researchers out there,

but they're not using their talents.

Right? For example, you know, if you're

in academia,

there's pressure to publish, right? And

if there's pressure to publish,

you think to yourself, okay, well, I

have this really cool idea,

but it might not work. It might be too

weird, right? It might be difficult to

get it accepted because I have to sort

of like sell the idea more.

Or I can just try this new positional

embedding, right? The problem is that

the current

environment both in academia and in

companies are not actually giving people

the freedom that they need to do the

research that they probably want to do.

I

>> mean there's also this interesting thing

that even in spite of great new research

I mean I was speaking to Seb Hoger and

he's got all of these new architectural

ideas and open AI aren't implementing

them. I mean Google are doing this

diffusion language model which is quite

cool. And I I'd like to know your

opinion on why that is. So there's a few

philosophies floating around like this

concept of a universal representation

that there are universal patterns and

the the transformer representations

resemble those in the brain. And it's

rather led to this idea of well we don't

need to use different architectures

because if we just have more scale and

more compute then all roads lead to

Rome. So why would we bother doing it

any differently?

>> There's actually better right? There is

actually already architectures that have

been shown in the research to work

better than transformers. Okay.

But not better enough

in order to move the entire industry

away from such an established

architecture where you're familiar with

it. You know how to train it. You know

how it works. You know how the internals

work, right? You know how to fine-tune

them. You have all this software is

already set up for training

transformers. fine gening transformers

inference.

So if you want to move the industry away

from that, being better is not good

enough.

It has to be obviously crushingly

better. Transformers were that much

better over RNNs. Okay, transformers

where you just applied it to a new

problem and it just was so so much

faster to train and you just got such

higher accuracy that you just had to

move.

And I think the deep the deep learning

revolution was also another example of

that, right? Where you had plenty of

skeptics and people were pushing um

neural networks even back then and

people are going, "No, we think symbolic

stuff will work better." But then they

demonstrated it as being so much better

that you couldn't ignore it. This fact

makes finding the next thing even

harder. Right? That's the gravitational

pole of always pull pulling you back to,

oh, okay, but a transformer is good

enough. And yeah, you made a cool little

architecture over here that yeah, it

looks like it's it's got better

accuracy, but OpenAI over here just made

it 10 times bigger and it beats that.

So, let's just keep going. May I also

submit that there could be an additional

reason which is you know I love that

fractured entangled representations

paper. Um there's there's this shortcut

learning problem and I think that

there's a little bit of a mirage going

on here and there there might be

problems with these language models that

we don't you know that we're not fully

aware of and there's also this thing

that we're seeing that we are starting

to bastardize the architecture. So we

know we need to have adaptive

computation for reasoning. We know we

want things like uncertainty

quantification and what we're doing is

is we're bolting these things on top

rather than having an architecture which

intrinsically does all of these things

that we know we need.

>> Yeah. And I and I think the the our

continuous thought machine is is an

attempt at addressing those um more

directly, right? Which which Luke will

be able to tell you more about later.

There's something still not quite right

with this the current technology, right?

I I think the the phrase that's becoming

popular is jagged intelligence, right?

That the fact that you can ask an LLM

something and it can solve literally

like a PhD level problem and then you

know in the next sentence it can say

something just so clearly obviously

wrong that it it's it's jarring, right?

And I think this is actually a

reflection of something probably quite

fundamentally wrong with the current

architecture. As amazing as they are,

the current technology

is actually too good. Okay.

Another reason why it's it's difficult

to move away from them, right? So

they're too good in in in the following

sense. And you you spoke about the fact

that we have these foundation models.

That's okay. so that we have the

foundation that we can do anything with

them. Yes, I think current neural

networks

are so powerful that if you have enough

patience and enough compute and enough

data,

you can make them do anything.

But I don't necessarily think that they

want to, right? we're sort of forcing

them like they're universal pro

approximators but I think there are

probably a space of you know function

approximators that will more want to

represent things in the way that a human

represents them. So there's actually

quite an obscure paper that is my poster

child for this. It's called intelligence

matrix exponentiation

>> and I think it was actually rejected.

So, you know, you can probably project

uh the image of a figure one, but

there's an image of it's solving, you

know, the classical spiral data set of

needing to separate the two classes in

the spiral.

>> Yes.

>> And it has the decision boundary for a

for both a classic RNN uh multi-layer

perceptron and a tanh multi-layer

perceptron. And you can see they both

solve it, right? Technically, they both

solve the problem because they they they

classify all the points correctly and

get a very good test score on this on

this very simple data set. And then they

show you the decision boundary for the

for the M layer that they built in this

paper and it's a spiral.

The layer represented the spiral as a

spiral. Sh shouldn't we should you know

if the data is a spiral shouldn't we

represent it as a spiral? And then if

you look back at the decision boundaries

for for the spiral and the classic relu

multi-layer perceptron,

it's clear that you just have these tiny

little peacewise linear separations.

Um, and that's what I mean. Yes,

if you know if you train these things

enough and you push these little

peacewise linear boundaries around

enough,

it can it can fit the spiral and get a

high accuracy. But there's no feeling

when I look at those that that image

that the relu version actually

understands that it is a spiral,

right? And when you represent it as a

spiral, it actually extrapolates

correctly because the spiral just keeps

going out.

>> You're touching on something fascinating

there because, you know, we were talking

about the need for adaptivity and um

adaptive computation. Um I'm really

inspired by Randall Bisreerero's spline

theory of of neural networks and we

we've had them on many times and you can

look on the TensorFlow playground. You

can look what happens when you have a

ReLU network on on this, you know,

spiral manifold. And, you know, you'd be

forgiven for thinking that these things

are basically a locality sensitive

hashing table, right? Because they they

do they they they they partition the

space and and they they can predict the

spiral manifold, right? But we want to

do something a little bit more different

than that. And it also comes into this

imposttor thing because just tracing the

spiral manifold

but not continuing the pattern there's a

big difference between that. So from an

imposttor perspective just just tracing

the pattern is not learning it

abstractly or constructively. Right? If

we learned it constructively so we you

know you speak about this in your paper

this complexification the abstract

building blocks and you can do adaptive

computation. you understand the spiral.

That means that with adaptive

computation, you can continue the spiral

and then you can update the model's

weights so it has adaptivity because

that's so important for intelligence. So

we know that we need models that can do

these things. But for some reason

they're they're so sick of fantic

they're almost better than an adaptive

intelligent system because they tell us

exactly what we want to hear. They seem

so intelligent, but we know that they're

missing these fundamental properties.

>> I'm still fairly skeptical when I see

video generation models.

You know, we went through a phase where

you could detect them because of the

number of fingers on somebody's hand,

right?

And yes,

with more data, with more compute, with

better training tricks, okay, they

submit. And now they usually do have

five fingers. But did we fix the problem

or did we just use more brute force to

just you know force the the neural

network to know it's five fingers where

something that actually had a much

better kind of representation space.

It's almost mad that it's controversial

to say that we should represent a spiral

like a spiral. But, you know, something

that could do that generally that if if

it represented a human hand the way

that, you know, maybe I represent a

human hand, then maybe it would be much

easier to count how many fingers are on

a on a hand. It's unfortunate

that they work so well. It's unfortunate

that scaling works so well because it's

too easy for people to just sweep these

problems under the carpet. You guys have

possibly created what I think might be

the best paper of of the year. This

could actually be the innovation which

takes us to the next step. And you did

you get the spotlight in Europe as well?

>> Yeah.

>> This year and congratulations on that.

So I think that's testament to how

amazing this paper is.

>> The CTM the continuous thought machine.

It's actually not that far out outside

of the local minimum that we're stuck

in. Right? It's not as if we went and

found this completely new technology.

Right? We took quite a um a simple

biologically inspired idea, right, of

these of of the fact that neurons

synchronize and not even necessarily in

a biologically plausible way, right?

Brains don't literally have all their

neurons wired together in a way that

they work out their synchronization.

But it's

it's the sort of research that I want to

encourage people to do. And the way to

sell it is quite easy. I think at no

point did we have to worry about being

scooped, right? That stress was taken

away from us completely. So, and there

was so there was no pressure to sort of

rush out with this with this idea

because we're like, well, there's

probably somebody else working on

exactly this. And I think the reason

that we, you know, we were able to get a

spotlight is because we're able to

create such a polished paper. We took

the time to do the science properly to

get the the base the baselines that we

wanted and do all the the the the tasks

that we wanted to try. Encouraging

researchers to take a little bit more of

a of a risk, right? To try these

slightly more speculative long-term

ideas, I think is is the sad thing is I

don't think it's necessarily a very

difficult thing to sell. And I want to

have the CTM as like a poster child of

it works, right? It was a bit of a risk.

We, you know, we didn't know if we were

going to find something interesting, but

you know, it was our first shot and we

did find something interesting and it

became a a successful paper. If if we do

find a system which can acquire

knowledge, design new architectures, do

the the open-ended type of science that

you're speaking to, can you see a future

where at some point the locus of

progress will be mostly driven by the

models themselves?

>> I think so. Whether or not that's going

to replace us completely, I go back and

forth on. Powerful algorithms are

finding uh helping us do research,

right? And I think it might just end up

being a more powerful version of that.

Right? So I I know the the AI scientist

that we we released, we showed that you

could actually go end to end, right? Go

from seeding the system with an idea for

a research paper and then just take your

hands off and just let it go.

Think about the idea, write the code,

run the code, collect the results, and

write the paper. uh to the point that we

were actually able to get it to um a

100%

AI generated paper accepted to to a

workshop recently, right?

But I think we did that to show that you

could do it right as a sort of

demonstration

in a real system. I think I would want

it to be much more interactive,

right? I would want to be able to seed

with an idea and then have it come back

with more ideas, have a discussion with

me, then go away to write the code. I

want look at the code and check it and

then discuss the results as they're

coming out. So that's the sort of

nearterm future that I that I would

envision or how I would like to do

research with an AI.

>> And could you introspect on that? Is it

because you feel we need supervision

because the models don't yet understand?

You know there's this path dependence

idea. So we need to do supervision

because we have the path dependence so

we can guide the generation of the

language models. Maybe in the future the

language models will just understand

better themselves. But there's also the

output dimension which is that we want

to produce artifacts that extend the

fogyny of human interest. We want it to

be human relevant.

>> Yeah. I I think it's more that you know

in that initial seed idea it's probably

impossible to actually describe exactly

what you want. It's exactly the same

with, you know, when I have an intern.

I can't just have an intern come into

the company and I go, I have this mad

idea and then just explain it to them

and then just leave them alone for 4

months.

There's a back and forth because I have

a particular idea that I want to explore

and I need to keep steering them in the

direction that I I you know that I had

in my my mind originally. So I think

it's more like that basically. You have

such a deep understanding. So you have

this rich provenence and history and

path dependence and that means you can

take creative steps intuitive steps for

you respect the fogyny. They respect all

of this deep abstract understanding that

you have and interns don't yet have that

>> but maybe AI models in the future will

have that.

>> Yeah sure. If they if they get to the

point where my inputs becomes

detrimental

then yeah that'll be a thing. It's kind

of like chess, right? There was a point

at which chess engine and human fusion

actually beat chess engines. That's not

that's not true anymore, right? Adding a

human into the mix actually makes the

the bots worse.

>> Oh, interesting. I wasn't aware of that.

>> Yeah. So, so what to do when that day

comes for AI scientists is a is a is a

broader discussion. I think

>> I think now is a good segue to talk

about this paper in a little bit more

detail. So this continuous thought

machines you were just pointing to it

before. Luke first first of all first of

all mate introduce yourself

>> and set this thing up for us.

>> My name is Luke. I am a research

scientist at Sakana AI [snorts] and uh

my primary sector of research is this

continuous thought machines. It's took

us somewhere in the region of about

eight months working on this project

with uh the whole team. Um I I did a lot

of the work uh but we also had a lot of

people in different areas and doing

different parts of it that I think an

8-month life cycle for a paper seems a

bit long for AI research at the moment.

Um but yes to the to the actual

technical points of the paper. So we

call it continuous thought machines. It

originally had a different name. We

called it asynchronous thought machines

before but uh every single time people

asked us what's the asynchronous part it

became a bit confusing. So continuous

thought machines basically depends on

three novelties. Uh the first one is

having what we call a internal thought

dimension and this is not necessarily

something new. It's related conceptually

to the ideas of latent reasoning. Uh and

it's essentially applying compute in a

sequential dimension. And when you start

thinking about ideas and problems uh in

this domain and in in this framework,

you start understanding that many

problems that look like intell or

solutions to problems that look

intelligent are often solutions that

have a sequential nature. So for

instance, one of the primary tasks that

we tested in the continuous thought

machines was this maze solving task. And

solving mazes for deep learning is is

quite trivial. It's really easy to do if

you make the task easy for machines. And

one of the ways to do this is you give

an image of a maze to a neural network

like a convolutional neural network and

it outputs a image uh same size of the

maze and it's uh zeros where there isn't

a path and there's ones where there is a

path. There's some really brilliant work

showing how you can train these in a

careful way and scale them up

essentially indefinitely. And this is

fascinating and uh really interesting

idea of how to solve this. However, when

you take that uh approach out of the

picture and you ask what is a more human

way to solve this problem, it becomes a

sequential problem. You have to say well

go up, go right, go up, go left,

whatever the case may be to trace a

route from start to finish. And when you

constrain that simple

problem space uh and you you ask a

machine learning system to solve it like

that turns out to actually get much much

more challenging. So this became our

hello world problem for the CTM and

applying an internal sequential thought

dimension to this is how we went about

solving this.

Uh two other novelties that we can touch

on and talk about. Uh we we sort of

rethought the idea of what neurons

should be. There is a lot of excellent

research in this world uh in cognitive

neuroscience particularly exploring how

neurons work in biological systems. And

then we get on the other side of the

scale how deep learning neurons work

which the quintessential example is a

relu. It's off or on in a sense. And

this very very high level abstraction of

neurons in the brains feels a little bit

myopic. So we approached this problem

and said well let's let's on a neuron by

neuron basis let this neuron be a little

model itself. And this ended up doing a

lot of interesting work on how to build

dynamics in the system. The third

novelty here is as I said before we have

this internal dimension over which

thinking happens. We ask the question,

well, what is the representation? What

is the representation for a biological

system when it's thinking? Is it just

the state of the neurons at any given

time? Does that capture a thought, if

you wish, if I can be controversial and

use the term thinking and thought and my

philosophy with this is no, it doesn't.

That the concept of a thought is

something that exists over time. So, how

do we capture that in in engineering

speak? We instead of measuring the

states of the model that is recurrent,

we measure how it synchronizes how

neurons synchronize in pairs along with

other neurons. And this opens up the

door to a huge array of things that we

can do with this type of representation.

>> You were talking about this um sort of

sequential nature of of reasoning and

devil's advocate. I mean there was that

anthropic biology paper and they were

talking about planning and thinking and

and they they were they were saying that

this thing is planning ahead because

because I think your system actually we

can say does planning it it's it's

actually different computationally can

you explain that

>> yes I think the boundary in terms of

computation from a a cheering machine

perspective if you wish uh is really

interesting because the notion of being

able to write your tape uh read from a

tape and then write again to be in a

Ting compute system ting complete system

is uh obviously an incredible idea that

has completely changed the world and I

think the primary difference with let's

talk about transformers versus what

we're trying to do with the CTM

is that the process that the CTM thinks

in we can apply that process that

internal process to uh breaking down a

problem. So the problem itself can be a

single there is a single solution to

this problem and you could do that in

one shot. You could as I explained with

the maze you could just process that in

one shot but there are certain phrasings

of problems that are real problems that

doing so becomes exponentially more

challenging. So in the maze task, a

really good example is that if you try

to predict 100 200 steps down the path

in one shot, no models that we could

train, not even our model could do that.

And we needed to actually build an

autocurriculum system where the model

first predicted the first step and then

when it could predict the first step,

then it we started training it on the

second and third and fourth step. And

the sort of resultant behavior of this

is where it gets interesting. One of the

one of the ways that I like to do

research and that I encourage people who

work with me to do research is

understand the if you wish the behavior

of a model. We're getting to a point now

where the models that we build are

demonstrably intelligent in ways that

keep surprising us and breaking that

down into a single set of metrics or

even a finite single metric about

performance seems maybe not to be the

right way to do it for me. and

understanding the behavior and the

actions that those models take when you

put them in a system and train them in a

certain way uh seems to reveal more

about what's actually going on under the

hood.

>> Very cool. And I think I didn't pick up

on this. So So you're you're doing a

fixed number of um steps so you have

like a a context window and did you say

that you've set that around 100 steps?

>> So for the for the maze task uh the

model always observes the full image at

every step. uh the CTM will absorb

observe the full image for argument sake

those images could be tokens from a

language uh the output of a language

model those uh inputs could be numbers

that the model has to sort whatever the

case may be it should be agnostic to

data that's how we've tried to build it

but in the maze task the model can

continuously just observe the data uh no

matter where it can look at the whole

image simultaneously but it uses

attention to retrieve information from

the data and it has let's call it 100

steps that it can think through. And

what we do is we pick up at some point

the model solves three steps through the

maze. So it says I'm going to go up, up,

and right. And then it's correct. But

then it makes the wrong turn. And at

that point, we stop supervision. We only

train it to solve the fourth step. So

one more than what it could. In

practice, we do it five, but the

principle holds. And when you do that,

it's a self bootstrapping mechanism. And

I think the uh intuitive listener will

understand how that extends to other

domains, other sequential domains for

instance like uh language prediction,

many tokens ahead, that sort of thing.

>> So I'm really interested in this idea of

adaptive computation. So I I guess the

first question is how sensitive was the

performance to the number of steps and

then the next question would be could

you have an arbitrary number of steps

which means that you know perhaps based

on uncertainty or you know some kind of

criterion you could do fewer steps and

then the final question is could you

have potentially like an arbitrary or

unbounded number of steps

>> yeah uh really super question I think

that

I think that I'll answer the uncertainty

question first about the sensitivity to

steps. So a very good example of this is

we just trained the model on imageet

classification and our last function is

quite simple. What we do is we run it

for for example 50 steps and we pick up

two points two distinct points. The

first one is where is it performing the

best i.e. where is the loss the lowest

and the second one is where is it most

sure or where is it most certain and

those give us two indices uh between 0

and 49 inclusive and we apply cross

entropy at both of those points we just

make the last the average of the cross

entropy at those points. So what this

does is it induces a behavior where easy

examples are solved almost immediately

in one or two steps whereas more

challenging examples will naturally take

more thinking and it enables the model

to use the full breadth of time that it

has available to it just in a natural

fashion without having to force it to

happen. So you've decided to model every

neuron as an MLP which is really

fascinating. Talk about that but also

there's this notion of synchronization

and I think you use the inner product to

determine the extent to which the

parameters are are synchronized and this

kind of unfills over over time as as the

driving force. Can you explain that in a

bit more detail?

>> Absolutely. I think it's a it's a good

point to explain the uh neuron level

models as we call them in the paper or

NLM first because it ties into this. So

you can imagine a recurrent system is a

state vector, a state vector that is

being updated from step to step. We

track that state vector and that state

vector unfolds and for each individual

neuron, each uh I neuron in the system,

we have a unfolding time series. It's a

continuous time series. Well, it's

discreet, but it's a continuous value.

And those time series define what we

call the activations over time. And

synchronization is quite simply just

measuring the dotproduct between two of

these time series. So you have a system

of d neurons and essentially you have d

over two squared different

synchronization pairs. So neuron one can

be related to neuron 2 by how they

synchronize and neuron one can also be

related to neuron 3 etc etc. The neuron

level models they function by taking in

a finite history like a FOQ of neuron of

activations coming in and instead of

being just a radio activation they use

that history as information to uh

process a single activation out and that

is what moves from what we call

pre-activations to post activations. And

the principle here is that this might

seem rather arbitrary and does it help

for performance? Turns out it does, but

that's not really the catch all solution

here. That's not what we're after. Uh

what we're after here is trying to do

something biologically plausible. Uh

find the line somewhere between biology,

which is how the brain implements things

in the biological substrate that we have

versus deep learning, which is highly

parallelizable, super fast to learn,

back propanable, all of the nice

properties of that that have got us this

far. and find a line somewhere where we

can take some sprinkling of biological

inspiration but still train it with deep

learning. And it turns out that neuron

level models is a nice interim that we

can do this with. The concept of

synchronization is applied on top of the

outputs of those neuron level models. So

on on this on the scaling the I think

the time complexity is quadratic in

respect of the dimension of the

synchronization matrix right and in your

paper you were talking about subsampling

to improve the performance but how how

did that affect the the stability and

the you know like were there any things

that that cost you doing that? Yeah,

it's a neat question. I think in terms

of stability, what's what we found was

kind of fun and this was a sentiment

that we had throughout the the

experiments that we ran with this paper

was it tended no matter what we tried it

on it it just kind of worked with all

spreads of hyperparameters. uh and this

the problems that you have with back

prop through time typically with

recurrence models like RNNs and LSTMs

it's a challenge and you run for many

internal ticks with the RNNs or the

LSTMs and the learning seems to break

down but the uh fact that we use

synchronization in some sense touches

all of the neurons through all of the

time so it really helps with gradient

propagation uh a nice interesting point

that's maybe a bit oblique to what you

asked about synchronization is we have a

system of d neurons and like I said

earlier there

d over two squared possible

combinations.

This essentially means that our

underlying state or underlying

representation to the system is quite a

lot larger than what you would get with

just taking those D neurons. And as to

what that means in terms of downstream

computation and performance and the

things that we can do with this is what

we're actively exploring right now.

>> You guys used an exponential decay rate.

>> You have the system that unfolds over

time. it would be maybe a little bit too

constrained if the synchronization

between any two neurons depended on the

same time scale. So for instance, there

are neurons in your brain that are

firing over very long time scales and

very short time scales. The way that

they fire together impacts other neurons

and causes those neurons to fire. But

everything in biological brains happens

at diverse time scales. It's why we have

uh different brain waves for different

thinking states for instance. Uh but

beside that point, what we do with the

exponential decay in the continuous

thought machines is it allows us for a

very sharp decay to say that these two

neurons that are pairing together, what

only really matters is how they fire

together right now. Right? But if we had

a very long and slow decay, essentially

that's capturing a global sense of how

those neurons are firing over an

extremely long period of time. So this

was essentially a way of us uh capturing

this idea of how different neurons could

maybe fire together very quickly and

other neurons can fire together very

slowly or not at all. And this lets that

representation space that I spoke about

that D over2 squ representation space

lets it again become more rich and we

can enrich that space with more subtle

tweaks to how we compute those

representations.

So we were speaking about this

yesterday, Luke, that um when folks

apply transformers to things like the

ARC challenge or things that need

reasoning. Um we need to do lots of

domain specific hacks. So the architects

who were the winners of last year's

challenge, they did um depth first

search sampling and some folks have been

experimenting with using language

representations or you know using um

DSLs and some part of this is to do with

the the the reachability um of language

right and and language is is quite dense

which means you can kind of

monotonically um increase but if I

understand correctly your system might

have some interesting properties for

reasoning and for discrete and sparse

domains and also for sample efficiency

because we want we want to build a

system that can actually do well on

things like the arc challenge. But can

you kind of explain in simple terms why

you think this architecture could be

significantly better than transformers

for doing those things?

>> I think a lot of the really fascinating

work in the last few years that I found

fascinating in the literature of

language models has been related to what

one can actually call a new scaling

dimension. I in some sense see continue

uh chain of thought reasoning as a way

of adding more compute to a system.

That's obviously just one small part of

what that really is and what that really

means. But I think it's quite a profound

breakthrough uh in some sense. Now what

we're trying to do is is have that

reasoning component be entirely internal

yet still running in some sort of

sequential manner. And I think that

that's rather important. And you spoke

earlier about Gemini's diffusion

language modeling and I think that there

are a lot of different directions that

are exploring this right now. uh I do

think that the continuous thought

machine with the ideas of

synchronization and multi-hierarchical

temporal representations gives a certain

flexibility on that space that uh other

people are not yet exploring and that

richness of that space being able to

project the next step to solve the arc

challenge and the next 100 the next 200

steps to be able to break that down into

a process that a model can then uh very

quickly search that process in its

highdimensional latent case becomes

something that feels like a good

approach to take.

>> Do you see any relationship between this

architecture and you know Alex Graves

neuro touring machine?

>> Yes, that's really interesting. Um I do.

I think that the one of the the most

challenging parts about uh working with

a neural neural touring machine is the

concept of writing to memory and reading

to memory because it is a discrete

action. Um and that's that has its own

challenges associated with it and

yes uh I wouldn't go so far as to say

that the continuous thought machine is

definitively nearing Tur incomplete but

the notion of [clears throat] the notion

of doing reasoning in a space that is uh

latent and letting that space unfold in

a way that is uh rich towards a

different set of tasks. And this this

actually brings me to a point that I

find quite interesting um that I'd like

to share with you. Consider again the

imageet task or any sort of uh

classification task. It's it's a nice

test bed. There are many images that are

really easy and there are many images

that are really difficult. When we train

for instance a vit or a CNN uh to do

this task,

it has to nest all of that reasoning in

the same space. It has to put all of its

decision-m process for a very simple

obvious cat versus some complex weird

underrepresented class in that system in

that data set and it has to nest it all

in parallel in a way that is we get to

the last layer and then we classify. Um

I I think breaking that down where you

have different points in time where you

can say now I'm done I can stop versus

now I'm done I can stop let you take a

data set or take a task and actually

naturally segment it into its easy to

difficult components. And I think we

know that curriculum learning and

learning in this continuous sense again

seems to be a good idea. It's it's how

humans learn. And if we can get at that

architecturally and just have that fall

out in a model, again, this seems like a

something worth exploring. Uh I'm not

sure if you know much about model

calibration and how neural networks tend

to be poorly calibrated.

>> Oh, go for it, Tommy. Um it's it's a bit

of an old finding, but if you train a

neural network for long enough and it

fits really really well and you've

regularized it regularized it really

really well, you'll find that the model

is unccalibrated, which essentially

means that it is very certain uh about

some components where some classes where

it's wrong and uncertain for some

classes where it's correct. Essentially

what you want for a perfectly calibrated

model is if it predicts a uh probability

that this is in class the correct class

with 50%. 50% of the time you want it to

be correct about that class and so on

and so forth. So a well-c calibrated

model if it's predicting a probability

of 0.9 that it is a cat then 90% of the

time it should be correct. And it's

actually turns out that most models that

you train for long enough get poorly

calibrated. And there are loads of post

hawk tricks uh to fixing this. We

measured the calibration of the CTM

after training and it was nearly

perfectly calibrated which is again a

little bit of a smoking gun that this

actually seems to be probably a better

way to do things. The flavor of this

kind of research is such that we didn't

actually go out and actually try to

create a very well-c calibrated model,

right? And we didn't even try to create

a model that was necessarily going to be

able to do some kind of adaptive

computation time, right? Um I was um a

very big fan of the the paper um uh yeah

adapted computation time was Alex Graves

was it? But that paper um it had a

massive amount of hyperparameter sweeps

in it because in that paper he needed to

have a loss on the amount of computation

that was being done.

>> Because anytime you try to do some sort

of adaptive computation time research,

what you're fighting is the fact that

neural networks are greedy,

right? because obviously the way to get

the lowest loss is to use all the

computation that you have access to. So

unless you had like an extra loss that

had a penalty that said okay actually

you're not allowed to use all the

computation that's and and very very

carefully balance loss that's when you

actually got the interesting dynamic

computation time behavior falling out of

the the model in that paper.

But was really gratifying to see with

the the continuous thought machine is

that because of the way that we set up

the loss that Luke described earlier,

adaptive computation times seem to just

fall out naturally.

So that's more the way that I think

research should go.

>> Okay? because we we don't actually have

like a specific goal

um like or a specific problem we're

trying to fix like that or something

we're trying to invent. It's more that

we have this interesting architecture

and that we're just following the

gradients of interestingness.

>> Yes. And on on that point, I I think

maybe the most exciting thing about your

paper is, you know, we were talking

about path dependence and um having this

understanding which is built step by

step, this process of complexification

and u I mean maybe this is this is um

apppropo in in the theme of world models

in in general and also active inference

and I say active inference in big quotes

because it's not KL Friston's active you

know maybe adaptive inference or

something like that but we want to build

agents that can continue to learn that

can update their parameters and most

importantly can construct path dependent

understanding and because it that's

completely different to just

understanding what the thing is. It's

how you got there is very important and

this architecture potentially allows

these agents using this algorithm to

explore trajectories in spaces find the

best trajectories and actually construct

an understanding which carves the world

up by the joints. Yeah, that's a that's

a really neat perspective. I haven't

actually thought about it like that, but

yes, I think um that particular stance

becomes really interesting when you

think about ambiguous problems because

carving the world up in one way is as

performant as carving it up in another

way.

>> Yeah. uh you know perhaps the

hallucination in language models is

carving the world up in some fine way

but it's just not performance in our

measure of this is hallucination and

actually that's not true but in some

other trace down the path of wanting to

carve the world up through a auto

reggressive generation of tokens you end

up in a different carve up of that world

and being able to train a model that can

be implicitly aware of the fact that it

is actually carving up the world in a

different way and and explore those

manners, those uh descents down the

carve up is something that we're after

and I think it's quite an exciting

approach to be trying to take a stance

of let's let's break up this problem

into small solvable parts and learn to

do it like that and how can we do this

in a natural way without too many hacks.

Yeah, it's something I've been thinking

about because um Shalet as much as I

love his measure of intelligence um

ideas is for him adapting to novelty is

getting the right answer and the reason

why you gave that answer is very very

important and in machine learning we

have this problem that we we come up

with this kind of cost function that

rather leads to this shortcut problem

but you know we could just build a

symbolic system we could be gi and and

we could say okay we need do this um

principled kind of construction of

knowledge maintaining semantics. Well,

we're not doing that. We're doing a

hybrid system. But there must be some

natural way of doing reasoning where in

spite of the end objective being this

cost function that because of the way

that we traversed these open-ended

spaces that we can actually have more

confidence mechanistically that we're

doing reasoning which is aligned to the

world. I think that's a great way of

seeing this particular uh avenue of

research and I think that obviously

we're not the only people thinking like

this and we're not the only ones trying

to do this. Um what we have is an

architecture that's amanable to it and

surprisingly so it wasn't again wasn't

the goal. It's not the goal to to do

this type of research. It's not the goal

to be able to break the world down into

these small uh chunks that we can

actually reason over in in a way that

seems natural. Instead, what we did was

pay respect to the brain, pay respect to

nature and say, well, if we build these

inspired things, what what actually

happens? What what different ways of

approaching a problem emerge? And then

when those different ways of approaching

a problem emerge, what big philosophical

and uh intelligence-based questions can

we then start to ask? And that's where

we're at right now. So it might feel at

times, especially for me, uh too many

questions and too few hands to answer

those questions. But I think the fun and

exciting thing and the encouraging thing

that I I can you know try to encourage

other younger researchers out there is

that uh you know do what you're passion

passionate about and figure out how to

build the things that you care about and

then see what that does. See what doors

that opens up and see how to explore

deeper into those domains.

>> We were talking about this yesterday,

weren't we? That you can think of

language as being a kind of maze.

>> Yes. like what is to stop us from taking

this architecture and building the next

generation language model with it. I

mean that that's honestly as you know

something that I am actively trying to

explore right now and uh yeah I think

the maze the maze task gets really

interesting when you add ambiguity to it

when there are many ways to solve the

maze and honestly this isn't something

I've tried yet and maybe it's something

I should try next week but it's

essentially you can imagine an agent or

the CTM in this case observing the maze

and taking a trajectory and surprisingly

we saw this we have a section in our

recently updated paper on AR archive the

final camera ready version of this paper

where we added an extra supplementary

section that is not in the main

technical report and that supplementary

section is basically hey we saw this

cool stuff happen and we list I think 14

different interesting things that

happened while we were doing the

research um that obviously didn't make

it into the paper but we wanted people

to know about these strange things that

happened and this is one of the strange

things where uh

we watched during training what was

happening. And at some time during

training, maybe halfway through the

training run, we could see what the

model would do is it would start going

one path in the maze and then suddenly

it would realize, oh no, damn, I'm

wrong. And would backtrack and then take

another path. But eventually it gets

really good and it does you some sort of

distributed learning in this because

it's got a attention mechanism with

multiple heads. So it can actually just

figure out how to do this pretty well

and refine its solution. But sometime

early on in the the learning it descends

multiple paths and comes back and

backtracks. We have a really fascinating

set of experiments that also showed and

this this we actually have some

supplementary material online showing

this where uh and I don't really know

what this says. It's kind of a deep

philosophical thing but if you're trying

to solve a maze but you don't have

enough time. Turns out that there's a

there's a foster algorithm to do it. And

this was this blew my mind when I saw

it. So if we constrain the amount of

thinking time that the model has but

still get it to try solve a long maze

instead of tracing out that maze, what

it does is it quickly jumps ahead to

approximately where it needs to be and

it traces backwards and it fills in that

path backwards and then it jumps forward

again leaprogs over the top and traces

that section backwards and then leap

frogs and it does this fascinating

leaprogging behavior that is based on

the constraint of the system. And again,

you know, this is just an observation we

made and what that means uh in a deep

sense and how it's related to uh giving

a model time to think versus not and is

it enough time to think? What happens?

What different algorithms does the model

learn when you constrain it in this way?

I find that quite fascinating and an

interesting thing to explore. Does it

tell us something about how humans

think? Does it tell us something about

how how we think under constrained

settings versus open-ended settings?

There's a number of cool questions you

can ask on this front.

>> You you guys are both huge fans of um

you know population methods and

collective intelligence and because we

can we can scale this thing up and we

can scale it out and what would it mean

to scale this thing out not only just in

a kind of um what do they call it

trivial paralization but in terms of

having some kind of weight sharing

between parallel models and so on. What

what what would uh what would that give

you potentially?

>> Uh this is this is a fun area of

research. So one of the active things

that we're trying to explore in our team

is uh concepts of memory long-term

memory and what what does this mean for

a system like this? So an experiment

that one can construct for instance is

to put some agents in a maze and let

them try solve this maze not not how we

did it in the paper but in a very

constrained setting where a agent can

only see maybe a 5x5 region around it

and we give that agent some mechanism

for saving and retrieving memories and

the task if you wish is to solve that

maze find your way to the end and the

model needs to learn how to construct

memory

such that it can get back to a point

where it's seen before and know I did

the wrong thing last time and go a

different route and you can then see

this with uh parallel agents in the same

maze with a shared memory structure and

see what actually happens when you can

all access that memory structure and

have a shared global like almost like a

cultural memory that we can access and

solve this global task by having many

agents trying to use this memory system

and I do think that memory is going to

be a very key element to what we need to

do in the future for AI in general.

>> So the subject of uh reasoning came up

just a second ago and I think there's a

perception that recently we made a lot

of progress in reasoning right because

it's actually one of the main things

that I think people are are working on.

We released a data set recently called

uh Sudoku bench and I was actually quite

happy to see it come up organically on

your uh podcast a few weeks ago.

[snorts]

>> Chris Moore,

>> right?

>> Yes. So,

I I wanted to tell you a little bit

about this benchmark because

I think I've been having a little bit of

issue promoting it because it doesn't on

the surface sound particularly

interesting because Sudoku has a sort of

a feeling that it's already been solved,

right? So, how interesting can a

collection of of Sudokus be for

reasoning? Exactly. We're not talking

about normal sedokus.

We're talking about variant sodokas. And

what variant sodokas are are usually

normal sedokus, right? So put the

numbers one to nine in the row, the

column, and the box,

but then literally any additional rules

on top of that.

And they're all handcrafted.

they all have extremely different

constraints.

Um constraints that actually require

very strong natural language

understanding.

So for example, there's one puzzle in

the data set where

it tells you the constraints of the

puzzle in natural language and then

says, "Oh, by the way, one of the

numbers in that description is wrong."

Right? So you have to be able to meta

reason about the rules themselves even

before you start uh solving the puzzle.

There are other puzzles where you have

um

a maze overlaid on the sodoku and the

rat has to work out a way through the

maze by following uh a path to the

cheese. But then there are constraints

on the path that it takes of like what

numbers and what they can be add up to.

Um it's difficult to really describe how

varied these these uh these variants

sedokas are and I think they're so

varied

that

if anyone was actually be able to beat

our benchmark they would necessarily

have to have created an extremely

powerful reasoning system.

Right now, the best models

um get around 15%, but they're only the

very very simplest and the very uh

smallest Sudoka puzzles in in the set.

Um we're going to be putting out a blog

post about um GPT5's performance and it

is a jump but it's still completely

unable to solve puzzles which are you

know humans can can solve.

And what I really like about this data

uh data set um and actually was the

catalyst for me creating it in the first

place it was that there was a there was

a quote by Andre Kapathy saying okay so

we have all this data it's from the

internet

um but what you really want right if you

wanted AGI you wouldn't want all of the

text that humans have ever created you

would actually want the thought traces

in their head as they were creating the

text, right? If you could actually learn

from that, then you would get something

really powerful.

And I thought to myself, well, that data

must exist somewhere.

My first thought was maybe

philosophy like uh you know there's

there's a type of philosophy where you

just write down your thoughts without

thinking like just stream of

consciousness.

I thought maybe that could work. Um, but

then when I wasn't thinking about it and

I was, you know, in my leisure time, I

was watching a YouTube channel called

Cracking the Cryptic.

>> Yes.

>> Where these uh these two British

gentlemen will solve these extremely

difficult Sudoku puzzles for you. Right.

Sometimes their their videos are four

hours long and they they're

professionals like this is their job.

And what was perfect I realized is they

tell you in agonizing detail

exactly what reasoning they used

to solve those particular puzzles.

Right? So we with their permission took

all of their videos which represents

thousands of hours of very high quality

human reasoning like thought traces

and

scraped them and made that available for

imitation learning. Right? Um we did try

to do this internally. Turns out that I

did a little bit too much of a good job

of really creating a very difficult

benchmark. Right. So, we're still trying

to get that stuff working and we'll

publish it that if we if we have some

success. Um,

yeah, I want to I want to really sell

the fact that this this reasoning

benchmark really is different, right?

Not only do you get something that's

super grounded, like you know exactly if

it's right or wrong, so you can do RL to

your heart's consent,

but you can't generalize very easily.

Each puzzle is deliberately designed by

hand to have a new and unique twist on

the rules called a breakin that you have

to understand. And right now, despite

all the progress we've made,

the current AI models can't take that

leap. They can't find these breakins,

right? They'll fall back to, okay, I'll

try no, I'll try five, I'll try six,

I'll try seven,

right? The the reasoning becomes really

boring and nothing like what you see in

the transcripts that we've we've open

sourced from this from this YouTube

channel. So I just want to put the

challenge out there right that this this

is a a really difficult benchmark and I

think progress on this benchmark will

really mean progress in AI generally.

>> Could you reflect a bit so after

watching this um Cracking the Cryptic

YouTube channel? How diverse were the

patterns? Because um Chris was saying to

me, oh you know these guys they go on

Discord servers, they get these creative

crazy ideas and I'm I'm obsessed. Maybe

it maybe I'm just being idealistic, but

I love this idea of there being a

deductive closure of knowledge, right?

That that there's this big tree of of

reasoning and we're all in possession of

different parts of the tree to different

depths. So the smarter and the more

knowledgeable you are, the deeper down

the tree you go. But in this idealized

form, there is one tree and all

knowledge kind of, you know, originates

or emanates from these abstract

principles. And we could in principle

build reasoning engines that could just

reason from first principles and it

might be um computationally irreducible.

So so you have to perform all of the

steps. And it feels like because we're

not in possession of the full tree. What

we need to do is kind of fish around. We

fish around to find Lego blocks. Oh,

that's a good Lego block. I can apply

that to this problem. And maybe that's

just what we need to do in AI for the

time being is is is we need to just

acquire as much of the tree as possible.

But could could we just do it all the

way down?

>> Yeah, fascinating question.

That tree is probably massive, right?

>> And as a human is solving these puzzles,

they're definitely learning in real time

and discovering new parts of this tree.

And it's it's sort of a meta task,

right? Because it's not just reasoning,

you're reasoning about the reasoning.

And I don't think we can. We have that

in AI right now. Because if you watch

the videos, they'll say something like,

"Okay, this looks like a parask or this

is a set theoretic problem or, you know,

maybe I should get my path tool out and

trace this this around." And of course

the professionals they do have this this

already massive collection of reasoning

Lego blocks as you say in their head. So

they'll recognize okay that type of rule

usually needs this kind of Lego block.

It's actually fascinating to watch how

good they are at just intuitively

knowing where you know someone like me

haven't solved as many needs to spend a

lot of time looking around like okay

maybe I should try this or maybe I try

this one. Um, but even they're not

perfect. So, you can watch them take a

certain kind of reasoning and start

building up. Okay, maybe we should solve

it like this and then go and know that

doesn't disambiguate it enough and then

backtrack and then go down another path.

Again, something that we do not see

current AI doing when they're trying to

solve uh this this benchmark. the the

tree is very big and I guess the

phoggenetic distance between many of

these motifs in in the tree is just so

large. So it's so difficult to jump

between and and and I and I think that's

why as a collective intelligence we work

so well together because we actually

find ways to jump to different parts of

the tree,

>> right? And I and I think that's probably

why the RL the the current state of the

RL algorithms that we're trying to apply

to this just isn't working because

in order to learn how to get these

breakthroughs to to understand what the

sort of nuance reasoning is to get these

puzzles,

you have to sample them. And that it's

it's such a rare space, you know, it's

it's such an specific kind of reasoning

that's required to get to the the

specific breakthrough that this kind of

[snorts] technique doesn't work, right?

And there's definitely a feeling in the

community like, okay, this is how you

just solve things now. Like we have RL,

yes, we can get these language models to

do what we want. It doesn't work for

this for this data set.

>> Guys, it's been an absolute honor having

you on the show. Just before we go, are

you hiring? Because we've got a we've

got a great audience of ML engineers and

scientists and um I think working for

Zakano would be the dream job.

>> That's very kind of you. Yes, we are

definitely hiring and as I said earlier

in this interview,

I honestly

want to give people as much research

freedom as possible. I'm willing to make

that bet, right? I think things that are

very interesting will come out of this.

And I think we've already seen plenty of

interesting things coming out of this.

So if you want to work on what you think

is interesting and important, come to

Japan.

>> And Japan just happens to be the most

civilized culture in the world.

[laughter]

>> All right.

>> It might be the opportunity of a

lifetime, folks. So um yeah, get in

touch, guys. Seriously, thank you so

much. It's been an honor having you both

on the show.

>> Thank you very much.

>> Thank you so much. It's been great.