Maybe impressive in one way, but I'm also pretty sure a simple n-gram Markov model (a la Niall on the Amiga) would have a lower loss on the test set.

Transformers don't scale down very well, in my experience - I used to train local models all the time as new ones were released, as I recall transformers were the first ones I couldn't get better results out of with my limited training data and GPU.

I love these counterfactual creations on old hardware. It highlights the magical freedom of creativity of software.

Just reminded me of the random sentence generator program on my Vic-20. I had changed most of the words to all the bad words a preteen could think up. So many laughs with the neighborhood kids.

Dissapointed - there was no 6502 code in the GitHub repo.

You can chat with the model on the project page: https://indiepixel.de/meful/index.html

It (v3) mostly only says hello and bye, but I guess for 25k parameters you can't complain. (I think the rather exuberant copy is probably the product of Claude et al.)

If you're running this in VICE, run it under the SuperCPU with warp mode on.

i hate ai, and i love the c64, but i'll allow it.

but can you make mac keyboards feel like a c64c?

A little disappointed to see PyTorch + Claude here. I was hoping for some "demo-scene" hand-crafted 6502 assembly, and hopefully training on the C64.

Eliza called, and asked if we saw her grand kids...

This would have blown me away back in the late 80s/early 90s.

(Or maybe not, if it doesn't perform better than random, I haven't actually tried it out yet. Some more examples would have been nice!)

I wonder how far you could push this while still staying period correct, e.g. by adding a REU (RAM Expansion Unit), or even a GeoRAM (basically a REU on steroids).

SuperCPU would also be an option, but for me it's always blurring the line of "what is a C64" a bit too much, and it likely just makes it faster anyway.

How does this compare to ELIZA?

Load”*”,8,1

Brings back memories

Ok now we need 1541 flash attention.

I'm not sure what the venn diagram of knowledge to understand what that sentence is suggesting looks like, it's probably more crowded in the intersection than one might think.

Interesting, I’ve always thought neural network progress was primarily bottlenecked by compute.

If it turns out that LLM-like models can produce genuinely useful outputs on something as constrained as a Commodore 64—or even more convincingly, if someone manages to train a capable model within the limits of hardware from that era—it would suggest we may have left a lot of progress on the table. Not just in terms of efficiency, but in how we framed the problem space for decades.

> 25K parameters is about 70 million times smaller than GPT-4. It will produce broken sentences. That's the point - the architecture works at this scale.

Since it seems to just produce broken and nonsensical sentences (at least based on the one example given) I'm not sure if it does work at this scale.

Anyway, as written this passage doesn't really make a whole lot of sense (the point is that it produces broken sentences?), and given that it was almost certainly written by an AI, it demonstrates that the architecture doesn't work especially well at any scale (I kid, I kid).