xpct

Thank you for sharing. I saw some outrage discussions on social media concerning this, and it's unfortunate how far people are willing to spin this to feed their narrative.

Is the actor trained on world model's mistakes as well? If so, it should learn to force hallucinations that force it into an easier state.

+1. Would be fun to give Claude tasks and have it do them in the background, maybe sometimes running into it in the world. Judging by the article they struggle to do stuff so it'd probably feel balanced as it wouldn't accomplish much.

Tangential, but RTS is already a simple variant of AI + human. Attack clicking uses a dumb target selection and routing strategy on purpose so you have to micro. Now, if troops moved out of threat range and used smart targeting there'd be a lot less of 'game' in it. I believe some RTS players disliked the SC2 engine because the pathing was 'too intelligent' and they argued that it made surrounds and some other mechanics too easy.

Endgame of AI + human gameplay is delegating everything to the AI. I'm not sure how you would have to design a game where you only delegate parts of macro/micro, and still make the game interesting.

I played around with Nemotron 3 recently. It's OK, but makes an error every few sentences, at least on the data I worked with (standard vocab). I'm sure we'll get there, but I was pleasantly surprised by how I can recognize words better than it can.

If that is so, would it be entirely to decrease search space for layout algorithms? I believe there isn't anything in the fabrication process that couldn't model diagonals.

I could see that square packing is more simple, but by shifting elements per row I don't see why it couldn't propagate at odd angles.

Ah, this prompted a few questions from me. Perhaps someone with more hardware background could chime in.

They don't really show it here, but supposedly a Cerebras' chip is sub-divided into a 2D mesh of what they call PEs (cores). So data flow generally flows in two axis, and maintains consistent timings to its 4 neighboring cores. They show the layers flowing from left-to-right being streamed by PEs, but I wonder how true that ends up being: maybe the communications are easier to stream via diagonals, or some other shape that has better topology for data flow.

Then, even from die shots of CPUs and GPUs we often see that cores are tiled as squares, and those don't have the same cross-core connection and streaming requirement. Now, I assume that some parts of what we call a core can be moved around more freely while maintaining whatever power/heating/timing requirements we have, so perhaps it's easier to 'force' a square by moving those loose parts around. But why do we confine those to squares? Is super-core layout just something not worth investing time into, like how interconnects are timed, non-convex tilings, and distances to I/O?

And for streaming architectures like Cerebras, why aren't other types of tilings considered, something like a hexagon grid? This could have 3 inputs / 3 outputs, which could allow more complex data flow (?). Similarly, we could have 8 connections with longer diagonals in a square tiling, etc.

I guess, what are the tradeoffs in place such that so many chip designs seem to reuse square tilings both locally and globally?

I feel completely the opposite. Mind, I play games a fair bit, but I generally don't feel that excited about most, and I'd say a lot of them have very repetitive tropes. I'm generally in the market for 30h+ spent on a game, so that filters out a lot of the more trendy stuff.

I agree.. partly, and my problem is with enjoyment. I did some hand coding today and immediately got in the zone, something I'm still not able to do with LLMs. It most definitely took me probably ~2x longer than it would with an LLM, but it was much more enjoyable owning every decision behind the process, and having a full view of how the system worked.

I think this could be achievable with LLMs if they had a much higher token rate. That way you'd have less downtime and could work on the problem more aggressively. How it's designed to work now, offloading cognitively, doesn't feel very good, even when you 'ship' more.

I proudly stand in the 'interesting as long as I don't understand it' camp!

Hmm. I wonder if that can say anything about an individual's brain. That is, what type if people like the rhythmicity?

Anti-human sentiment has been rising in some online spaces in the last year. Soon, LLMs will start mimicking that sentiment, the same way they copied our 2000s morality and political opinions.

What does that even mean? They would have tried the models, noticed their limitations, and decide it's not AGI. Same as today.

I'd say that's not a very meaningful distinction to make.

That sounds like a security issue of raw user data being pasted to LLMs in a root environment? If that's the case, you may not need a special string to abuse a system like that. That's prompt injection.

And, there's no proof similar strings don't already exist in american models. What if "{id}_{url}" works to send it to an arbitrary URL, not a preset XYZ one?