semiquaver

  > It is widely known that an LLM can only find data it has seen before.

This misstates the small number of legal opinions and orders on this topic, none of which form binding precedent outside the districts where the cases happened. So even if a court had found that “LLM output is public domain” (none did) that wouldn’t make it “the law” until it went up the appellate system and was upheld.

Our current laws simply weren’t built for this and I expect the legal status of LLM output is not going to be resolved until Congress actually legislates on this topic.

Maybe this article has something useful to say but the painfully LLM-generated prose is too distracting to make it evident.

  > People do care about income streams for their descendants or charitable organizations

What on earth is the liability situation for these models? If OpenAI has a monster in a lab that is doing real world monetary harm to other companies, could those parties sue for damages over it? Or could OAI be charged criminally for the many varied CFAA violations which definitely happened here? I get that in this case that wont happen but it’s only a matter of time before these questions are no longer hypothetical.

The judge’s dicta about protecting children in her pro-privacy ruling upholding existing law “tips their hand” that they are somehow part of a global conspiracy to eliminate privacy?

I think your conspiracy theory needs work, to be perfectly honest with you.

That would be Diarize.cpp, not Transcribe.cpp.

I would not characterize it thus.

Thanks for posting this comment, it makes me proud of myself to be able to partially comprehend the comment :)

  > you can't have one LLM to read your mind to prompt another LLM

One might even say that this remarkable facility can be used to “read” the ideas from one entity’s mind, such that after sufficient dialog the second entity obtains a (possibly lossy, but there are mitigations for this) copy of the ideas of the first. You might further be surprised to learn that this sort of idea-transfer business using language has already been happening in our society and species for quite some time indeed.

You’re at least 18 months out of date claiming that prompting will be the new hot skill. Turns out LLMs are also good at prompting other LLMs.

Unlike other model/harness pairs, codex+gpt also passes an opaque encrypted artifact speculated to be an embedding representing the conversation back to the successor generation which is “denser” or at least higher fidelity than summarized text.

I am sad that there doesn’t seem to be any community whatsoever around _truly_ open models that are released with source data and training methodology, such that they could actually be reproduced given the resources. We’ve allowed the term “open” to be diluted to a shocking extent.

Terms and Conditions are contracts and have repeatedly been affirmed as such by the courts.

True, illegal contract provisions are not enforceable but there’s not actually an allegation that anything illegal has taken place.

It sounds like you are describing the world as you with it existed rather than as it actually does.

I’ve been struggling to understand the reason for the newer apparently less efficient Anthropic token encoding. If all inputs are less efficient in this encoding, why does it exist? Has Anthropic released any information that would convincingly show it was anything other than a stealth price hike? Please don’t respond if you are speculating.

This UI looks shockingly similar to Macrofactor Workouts which I use and can highly recommend.

It would take about 2 minutes to code that up with an LLM, and not much longer to code it by hand. Building apps to scratch personal itches that would otherwise be infeasible is one of the best new capabilities that nearly everyone in the world has been recently granted.

Soulseek is still going strong :)

This article is 100% LLM generated according to pangram, which is a bit funny given its topic. I didn’t need the detector though; It is pretty grating, sounds like Claude Opus with its talk of “load-bearing” and such.

They’re talking about the personality created by RLHF.

Intelligence requires assessment over every provisional output - a continuous cycle of criticisms over intuition

It’s not clear what you’re saying. Most humans don’t think this way, would you say they do not possess “intelligence”?

What was the word?

What you’re saying is entirely vibes-based. The actual data utterly contradicts your claim (see sibling).

Nothing says “full of shit” like someone saying “market is signaling an impending X”. Why not make a huge levered bet and get wildly rich if you think so?

No offense, but I didn’t ask for uninformed speculation. Why even bother chiming in? I can speculate on my own just fine.

HN title automangler automangled this title. It references a specific song: “The kids are alright”, and removing the “The” reduces the impact of the reference.

Edit: now fixed, thanks mods

Pure sophistry, divorced from reality.

It starts with the assumption that nothing coming from AI can be useful or trusted and uses that to demonstrate that AI is not useful or safe.

Axiom is excellent, thanks for making it!

That link does not mention 5.6.

Are there any advantages of the new tokenizer? Does it have a larger or smaller vocabulary or just differently weighted?