Ah, never mind, I believe I found the relevant hyperparameters here: https://github.com/adamimos/epsilon-transformers/blob/main/examples/msp_analysis.ipynb
In particular, the stuff I needed was that it has only a single attention head per layer, and 4 layers.
No, the actual hidden Markov process used to generate the awesome triangle fractal image is not the {0,1,random} model but a different one, which is called “Mess3” and has a symmetry between the 3 hidden states.
Also, they’re not claiming the transformer learns merely the hidden states of the HMM, but a more complicated thing called the “mixed state presentation”, which is not the states that the HMM can be in but the (usually much larger number of) belief states which an ideal prediction process trying to “sync” to it might go thru.
Can you share the hyperparameters used to make this figure?
Okay my computer right here has 10^13 bits of storage and without too much trouble I could get it to use all that memory as a counter and just count to the highest value possible, which would be 2^(10^13) or in other words much much longer than the age of the universe even at a fast clock speed.
Now technically yes, after it got to that 2^(10^13) value it would have to either halt or start over from 0 or something… but that seems not so practically relevant to me because it’s such a huge integer value.
I haven’t dived into this yet, but am I right in guessing that the gist is exactly like a way more fleshed-out and intricate version of Hofstadter’s “superrationality”?
I’m trying to read this post now but it looks like a bunch of images (of math) are missing. Does that match what others see?
The Bitter Lesson applies to almost all attempts to build additional structure into neural networks, it turns out.
Out of curiosity, what are the other exceptions to this besides the obvious one of attention?
Upvoted because this mentions Nonlinear Network.
Some of your YouTube links are broken because the equals sign got escaped as “%3D”. If I were you I’d spend a minute to fix that.
Have you read https://www.lesswrong.com/posts/5wMcKNAwB6X4mp9og/that-alien-message yet?
I had some similar thoughts to yours before reading that, but it helped me make a large update in favor of superintelligence being able to make magical-seeming feats of deduction. If a large number of smart humans working together for a long time can figure something out (without performing experiments or getting frequent updates of relevant sensory information), then a true superintelligence will also be able to.
Hilarious… I fixed my error
Reminds me of this from Scott Alexander’s Meditations on Moloch:
Imagine a country with two rules: first, every person must spend eight hours a day giving themselves strong electric shocks. Second, if anyone fails to follow a rule (including this one), or speaks out against it, or fails to enforce it, all citizens must unite to kill that person. Suppose these rules were well-enough established by tradition that everyone expected them to be enforced.
Keenan Pepper
What I gather from https://www.lesswrong.com/s/HzcM2dkCq7fwXBej8 is that it’s sort of like what you’re saying but it’s much more about predictions than actual experiences. If the Learning Subsystem is imagining a plan predicted to have high likelihood of smelling sex pheromones, seeing sexy body shapes, experiencing orgasm, etc. then the Steering Subsystem will reward the generation of that plan, basically saying “Yeah, think more thoughts like that!”.
The Learning Subsystem has a bunch of abstract concepts and labels for things the Steering Subsystem doesn’t care about (and can’t even access), but there are certain hardcoded reward channels it can understand. But the important thing is the reward signals can be evaluated for imagined worlds as well as the real immediate world.
Actually I would still really appreciate the training hyperparameters like batch size, learning rate schedule...