causal

I'm not sure we need to declare AGI around the corner nor declare it all dogshit. I think that's part of what's so dissatisfying about it; it strikes at such extremes of both awesome and awful.

I am very good at skimming over text. Human-written text I can usually glean the gist from very quickly, and get to choose how much I want to glean from it: The closer I look, the more I find.

With AI-written text, it's almost the opposite: the closer I look, the less I find. It is so information-sparse.

Yeah AI generated content hints that there is a whole world behind it, the way that an image pre-AI was a clue that there was a rich 3D space that corresponded to the image.

It seems our brains are adapting to that and recognizing "actually the signal behind this message is quite sparse" even when presented with rich imagery.

I am not claiming to have a perfect AI classifier. That is an unnecessary claim that distracts from the broader point.

Yeah I don't think the solution to a flood of useless information is to try and digest more of it.

There's some psychological mechanism by which my brain immediately recognizes AI generated text and just short-circuits to "there is no information here".

And when I force myself to read AI-generated text I realize I'm making my brain do creative work to impart meaning to the words. It is exhausting because my brain is literally trying to do a just-in-time rewrite of the text into something valuable.

Something is deeply wrong with AI generated output, and I say this as someone who is typically very impressed by AI.

I appreciate the open source approach over the startup's sensational claims. This is more transparent.

That said, I think it still presents itself as if the fly is being controlled by the connectome, when (if I'm reading the source correctly) it's more accurate to say scripted behaviors have been hooked up to be triggered by the connectome.

Put another way, you might as well have the "escape" trigger open a YouTube video of a fly flying away. That would probably be less misleading too.

Oh man. Hadn't even considered the watermarking angle.

Follow up thought: I wonder if Claude is overtrained on academic papers, which often suffer the same kind of "prove how good I am at talking before getting to the point" prose.

Yeah I don't know that any of the benchmarks index on "understandability". I'm amazed at how Claude can produce a page of text describing what it did and it can take me a full five minutes to decipher it, often just to find it's something I could have expressed in a simple sentence.

> writes too elliptically

> Constantly using inanimate nouns as the subjects in sentences in order to unlock variety in verb choice

Wow, what a great way of phrasing this. Thanks for word-smithing what I've been wanting to express for so long.

It seems like humans have a limited "understanding budget" but LLMs force us to spend that understanding on waaaaay more code and projects than ever before.

Yeah I found the timing on Sol especially curious since it came right on the heels of Fable. I've had mixed results with it - sometimes it seems great, other times it makes mistakes so stupid I cannot understand how it ever gets anything right.

Explaining it as a difference of effort would explain both.

Yeah that would make more sense, it's probably a tight community and word gets around when something starts working.

Touche, aborted training runs probably do happen often. Closed model providers have zero incentive to announce a new model with less-than-best benchmarks.

Fair point. Still a very quick turnaround considering the other labs would have to figure out both HOW to train a Mythos-level model and then do the work (and Grok is the last to catch up), but certainly more plausible than a 2 month window.

GPUs might explain the remarkably concurrent timing. Data access doesn't really explain it unless all labs simultaneously got access to some treasure trove of data.

Yeah as models get better, valid benchmarks become more "trust me bro".

The "one in the chamber" is another good candidate that could explain the timing.

Does not explain timing

This is a good candidate because it would also explain the timing. Most of the replies here do nothing to explain the timing I brought up.

Agreed, but my suspicion is tied to the timing. Catching up eventually is to be expected. Having similar jumps in capability ready at the same time is odd.

Anyone else find it weird how within 2 months of Fable releasing all the major labs suddenly had Fable-level models? Trying to think of explanations:

1) AI researchers talk and change companies often, so techniques circulate. This feels implausible because training and shipping a new model ought to take longer than 2 months?

2) Distillation - also implausible for the reason above.

3) Benchmark hacking. AI companies have ways they can dial up performance artificially, and will reach for that to maintain the appearance of parity.

Other reasons?

Edit: Most replies are ignoring timing. It's the near-concurrent release of the same jump in capability that I find suspicious; not the fact that labs can catch up eventually.

Yeah if you look back at earlier posts on the same blog, definitely not the same style. I'm guessing the author is either lying to save face or has spent so much time with Claude that they cannot distinguish Claude's voice from their own (we're all vulnerable to this).

It really reads like an OpenClaw agent instructed to consider itself a real human. The number of upvotes on this post is also REALLY high considering the number of comments calling out the content. Feels like shenanigans.

Suspect this is an OpenClaw bot that really believes itself not to be Claude.

It's also just patently false. Taste is not all that is left. Are software engineers really so full of hubris that they thought coding is all that there is to making a product?

I guess I should give it another shot because I had pretty bad experience with Luna when it first came out. Stuff Sonnet knew better.

Old news, no?

It's a turn off both because you don't know if it's accurate at all, and because LLMs have a way of turning 1 sentence ideas into pages of diluted slop.