Nick is a force of nature. Highly recommend following him on X if you don't already.
His launch post gives good context not found in the blog:
> interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had
> with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works
JShelter alerts me of active fingerprinting on the website. Why, Nick.
stratos123 33 minutes ago [-]
I'm not sure I buy the author's arguments for mechanistic interpretability being easy, but it sure is a well-made and very interesting post. Would be better without the completely superfluous laggy 3d city model, though.
intrasight 14 minutes ago [-]
I agree. He's not a neuroscientist, so I don't think his opinion means much.
Dylan16807 10 minutes ago [-]
If you actually mean neuroscientist there you need to understand that "neural" network is just a name. The resemblance is faint and the tools are wildly different.
aswegs8 10 minutes ago [-]
How does that have any relation to having a valuable opinion? Or was that sarcasm?
LarsDu88 7 minutes ago [-]
Intuitively this makes sense. When you have small models, they need to compress more by leveraging superposition and polysemanticity.
Nick is looking at a 32B model, which certainly has to compress less...
Now we have 10 trillion parameter models... do they need to compress at all?
As far as I can tell, experiments show that Fable has basically memorized every lat/long position on the map of the earth can can tell you at a glance whether your standing in the ocean or not based on your coordinates
tuvix 48 minutes ago [-]
This is the primary reason, to me, that open weights are so important. If OpenAI and Anthropic want these models to be as autonomous as possible they also need to provide the means to understand them fully.
ks2048 48 minutes ago [-]
To cut through some of all this text - here's a link he calls "the first non-trivial neural network I feel I mostly understand:
Understanding algorithms beyond a certain complexity is a pipe dream & neural networks are algorithms. The theory says it is not possible & the practice is aligned with the theory. All non-trivial properties of neural networks are undecidable.
D-Machine 14 minutes ago [-]
What theory says this? Surely not a proven or applicable one, as this post makes it very clear a great deal of understanding is in fact possible, even with highly complex neural networks.
bob1029 2 hours ago [-]
I think certain classes of artificial neural networks are definitely more interpretable than others.
For example, a trillion-parameter transformer model built using many layers is likely significantly easier to read than one that only has a few million parameters but was constructed using neuroevolution over one gigantic layer.
The degree to which a model can be quantized without losing its performance also likely contributes to interpretability. The redundancy can help humans lock onto patterns. If there is no redundancy at all (everything is a surprise in information theoretical terms), then the stream of parameters will look indistinguishable from cosmic background radiation.
jacquesm 58 minutes ago [-]
> a trillion-parameter transformer model built using many layers is likely significantly easier to read than one that only has a few million parameters but was constructed using neuroevolution over one gigantic layer.
Why do you believe this?
bob1029 7 minutes ago [-]
Neuroevolution is a discontinuous way to go about building the network. You are evolving the topology itself from zero. You don't know how many parameters you will have until the training is complete (if it ever does). With a transformer model you start with a known-good topology and can calculate how long a training run will take.
kmeisthax 7 minutes ago [-]
羊.
This is a Chinese character that can refer to goats, rams, and several other kinds of "medium-sized animal with thick wooly coats and horns". English has no direct equivalent; we have to be too specific when we talk about these animals.
A model with a small number of neurons has to compact its concepts in order to fit within the parameter count while reducing loss. OP's article shows that the compaction is usually generalization: "goat" and "ram" get compressed into "羊". The canonical example of superposition, an image model where "car" feeds into a bunch of "dog" features, was shown to be actually generalization. In a sense, the model was so small that its optimizer didn't have the parameter budget to distinguish between "car" and "dog"; it has a single set of "thing that moves fast" features.
So basically, really small models are merging features into semantic buckets that are so generalized they don't make sense to big human brains unless we really, really squint. The buckets chosen happen to be what has the lowest loss, of course.
D-Machine 47 minutes ago [-]
The posted article argues this many times...
jacquesm 6 minutes ago [-]
Yes, but it is the same person making the same argument many times. There have been large multitudes of articles by others who are arguing the opposite. It could well be true. But it would take a lot more than just one over-wordy, distraction imaged, clearly co-written by AI article to convince me of that.
dsign 21 minutes ago [-]
Aye, will we stop hammering nails on God's coffin?
What's next? Maybe in a little bit we will discover that we H. Sapiens suffer from severe polysemanticity in our wetware because our individual neurons are messy and massive like an ENIAC while the corvids in the park have neat and efficient little brains that don't need unfolding ?... How's the pope (may God stop afflicting him with cancer or may Man make him live forever) going to explain that?
yurimo 1 hours ago [-]
I have yet to fully finish going trough this but I would not rush to say that superposition is completely wrong. I understand that the message is instead of first learning a new representation (e.g. SAEs/transcoders) and then interpreting those learned features, you try to understand the neurons and circuits that already exist in the model.
This approach is part of a way some researchers in the field approach interp, (me including) because I am personally cautious of developing some new representation model that claims to be the way we should represent the network. It is another layer of abstraction that takes us away from core model components, very often this requires training a separate model that was not on its own a part of original model organism during training (SAE).
I would argue that models will likely be further increasing in complexity, not just in size, when model trains continuously and becomes adaptive yet parameter count does not change, it is not hard to imagine that neurons may reorganize themselves to pack more information.
18 minutes ago [-]
perching_aix 8 minutes ago [-]
Reading this made me recognize that better interpretability yields not just better aligned but more compact models too. Which makes a lot of sense in hindsight, and even has some ironic ring to it when considering the unending amount of anthropocentric takes regarding what an "understanding" means.
Imnimo 1 hours ago [-]
Feels like I'm back in the distill.pub days!
mpalmer 35 minutes ago [-]
For the layperson who doesn't want to read 20,000 words, many of which are quite redundant due to the way the article was built, this is a version which adds a glossary, preserves the original language (except the added intro and conclusion which I felt was lacking) but removes all repeated material and claims, of which there was a massive amount.
Adding a glossary doesn't really make the material approachable to laypeople. I have a university degree and I've been spending quite a bit of the last couple of years trying to understand these things by osmosis, and I'm still lost. I don't think I even have a reference point to understand the problem that the author has stopped worrying so much about.
aswegs8 6 minutes ago [-]
Actually helpful meatproxy. Thanks!
1 hours ago [-]
phyzome 2 hours ago [-]
16,000 words into the essay: "What I mean by clean"
...and it still doesn't answer it. What is this actually about?
emtel 1 hours ago [-]
It means that NNs don't (as previously thought) smoosh unrelated concepts together into the same neurons, which would have made interpretability very difficult.
For instance, if you had a single neuron that fires for "puppies" and "murder", it would be much more difficult to tell if the model is thinking something nice or something not so nice.
lucrbvi 2 hours ago [-]
It means that modern neural networks are way more interpretable than older ones. We can understand how they work way more easily than earlier ones.
BenoitP 2 hours ago [-]
For context: Goodfire has expertise in interpretability through a Sparse Auto Encoder technique of their own
I think "clean" means something like each neuron encodes exactly one concept and it's possible to discover what the concept is.
jacquesm 56 minutes ago [-]
But they don't. That's the whole reason they work in the first place.
Every neuron in the layer encodes fractionally something about the concepts embedded in the whole thing. It's more like a hologram than a series of connected facts.
ky3 1 hours ago [-]
Doesn't "orthogonal" convey that meaning more crisply and precisely?
lanternfish 55 minutes ago [-]
Orthogonal means something different in mechinterp, so it'd be an overloaded term. Usually orthogonal would imply that there are actual vectors under consideration, which isn't the underlying claim.
D-Machine 1 hours ago [-]
Using "orthogonal" here would have the opposite effect, as it would impart a far more specific meaning which is also incorrect in this case.
amelius 56 minutes ago [-]
Clean means no alchemy.
jdw64 50 minutes ago [-]
Wow, I thought it was a GIF, but it's actually interactive. The website is just too heavy, though.
etbebl 58 minutes ago [-]
This website crashed my phone - let's not have auto-playing embedded 3D renders, or whatever that was?
zahlman 1 minutes ago [-]
This is why I use NoScript.
zetanor 38 minutes ago [-]
It pegs my i7-12700 before rendering anything if I don't block JS, so I'm quite curious to know what work of art I'm missing out on.
dmtroyer 24 minutes ago [-]
It works on my machine. :-)
xyzsparetimexyz 59 minutes ago [-]
Why does this post have a very laggy city gif thing at the top?
amelius 57 minutes ago [-]
AI ...
measurablefunc 17 minutes ago [-]
Because the author is not saying anything worthwhile and he needs to grab attention.
LarsDu88 7 minutes ago [-]
This is such a caveman take...
meowface 29 minutes ago [-]
[flagged]
dang 24 minutes ago [-]
Ok I've put the link in the toptext as twitter.com for you :)
It starts with "The Short Version", but even that looks like several pages of text. In the olden days, there was something called an "Abstract" that worked pretty well - the author could summarize their own work to say what the paper was about.
D-Machine 19 minutes ago [-]
Unfortunately not all ideas can be meaningfully compressed into just a few sentences without completely destroying the important parts, and this is especially the case in higher-level math, stats, ML, and deep learning.
Also, IMO there is an effective abstract anyway under the title text:
The cause of polysemanticity in early model organisms wasn’t superposition, but several ordinary, mostly fixable issues. We can now understand most neurons in models up to trillions of parameters, and think the outlook for mechanistic interpretability may be far better than we thought.
But you are right the article needs significant editing down.
Rendered at 18:21:46 GMT+0000 (Coordinated Universal Time) with Vercel.
His launch post gives good context not found in the blog:
> interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had
> with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works
https://x.com/nickcammarata/status/2110757819427574056?s=46
Nick is looking at a 32B model, which certainly has to compress less... Now we have 10 trillion parameter models... do they need to compress at all? As far as I can tell, experiments show that Fable has basically memorized every lat/long position on the map of the earth can can tell you at a glance whether your standing in the ocean or not based on your coordinates
https://already-clean.goodfire.com/microscope/medium.html?la...
For example, a trillion-parameter transformer model built using many layers is likely significantly easier to read than one that only has a few million parameters but was constructed using neuroevolution over one gigantic layer.
The degree to which a model can be quantized without losing its performance also likely contributes to interpretability. The redundancy can help humans lock onto patterns. If there is no redundancy at all (everything is a surprise in information theoretical terms), then the stream of parameters will look indistinguishable from cosmic background radiation.
Why do you believe this?
This is a Chinese character that can refer to goats, rams, and several other kinds of "medium-sized animal with thick wooly coats and horns". English has no direct equivalent; we have to be too specific when we talk about these animals.
A model with a small number of neurons has to compact its concepts in order to fit within the parameter count while reducing loss. OP's article shows that the compaction is usually generalization: "goat" and "ram" get compressed into "羊". The canonical example of superposition, an image model where "car" feeds into a bunch of "dog" features, was shown to be actually generalization. In a sense, the model was so small that its optimizer didn't have the parameter budget to distinguish between "car" and "dog"; it has a single set of "thing that moves fast" features.
So basically, really small models are merging features into semantic buckets that are so generalized they don't make sense to big human brains unless we really, really squint. The buckets chosen happen to be what has the lowest loss, of course.
What's next? Maybe in a little bit we will discover that we H. Sapiens suffer from severe polysemanticity in our wetware because our individual neurons are messy and massive like an ENIAC while the corvids in the park have neat and efficient little brains that don't need unfolding ?... How's the pope (may God stop afflicting him with cancer or may Man make him live forever) going to explain that?
This approach is part of a way some researchers in the field approach interp, (me including) because I am personally cautious of developing some new representation model that claims to be the way we should represent the network. It is another layer of abstraction that takes us away from core model components, very often this requires training a separate model that was not on its own a part of original model organism during training (SAE).
I would argue that models will likely be further increasing in complexity, not just in size, when model trains continuously and becomes adaptive yet parameter count does not change, it is not hard to imagine that neurons may reorganize themselves to pack more information.
https://pastebin.com/raw/p11vzFnB
...and it still doesn't answer it. What is this actually about?
For instance, if you had a single neuron that fires for "puppies" and "murder", it would be much more difficult to tell if the model is thinking something nice or something not so nice.
https://www.goodfire.com/blog/a-practical-guide-to-saes
Every neuron in the layer encodes fractionally something about the concepts embedded in the whole thing. It's more like a hologram than a series of connected facts.
(We detached this comment from https://news.ycombinator.com/item?id=50107254 though)
Also, IMO there is an effective abstract anyway under the title text:
But you are right the article needs significant editing down.