There are more breakage in rust due to the type-system-related semantics, but ideally a C dev would also want their system to break if the semantics aren't right. So this is a criticism on C..?

So following this argument, they don't want Rust because C falls short? Nonsense.

edit: The speaker did mention that they didn't want to force limited use on the base APIs, but that for a great deal of their usage, they could have determined fixed semantics, and make intermediary APIs for it. So this was not about limiting the basic APIs.

Maybe it's better to carefully control and explain the talk progression. Selectively removing old prompts (adapting where necessary) - which also reduces the context - results in it not having to "bother" to check for inconsistencies internal to irrelevant parts of the conversation.

Eg. asking it to extract Q&A from a line of text and format it to json, which could be straightforward, sometimes it would wonder about the contents from within the Q&A itself, checking for inconsistencies eg:

  - I need to be careful to not output content that's factually incorrect. Wait but I'm not sure about this answer I'm dealing with here..
  - Before the questions were about mountains and now it's about rivers, what's up with that?
  - etc..

I think the real market for this is for local inference.

Since the Rust function has implicit/automatic behavior depending on how it's state is and how it's used by the callsite, and since the C one doesn't have any implicit/automatic behavior (as in, separate/explicit lifecycle calls must be made "manually"), I don't even see the reason for them to have the same name.

That is to say, having the same name would be somehow wrong since the functions do and serve for different stuff.

But it would make sense, at least from the Rust site, to have documentation referring to the original C name.

If you rearrenge all pixels from square-sized images using the Hilbert curve, you should end-up with pixels arranged in 1D, and that shouldn't be much different from "word tokens" that LLMs are used to deal with, right? Like a LLM that only "talks" in pixels.

This would have the benefit that you may be able to use various resolutions during training with the model still "converging" (since the Hilbert curve stabilizes towards infinite resolution).

I'm not sure if the pixels would also need to be linearized, then maybe it could work to represent the RGB values as a 3D cube and also apply a 3D Hilbert curve on it, then you would have a 1D representation of all of the colors.

I don't really know the subject but I guess something like that should be possible.

This new stance is marvellous and examplifies how free people should interact.

So either the copyright laws are "in practice" more getting in the way than actually helping anyone, or Amazon is missing punishment for abusers of copyright-related stuff.

Amazon allowing good-ratings buying, personal info scavenging etc puts Amazon's credibility in question, and I believe that this will eventually take it's toll.