roywiggins

I think if you asked "are humans usually more interested in being right or having friends", many- possibly most- people would say the latter.

The problem with asking these models stuff like "what do you think about humanity" is that, even if it can synthesize an insight about humanity, it's just as likely it's regurgitating some opinion it's already seen. Which, given that this "controversial idea" is hardly new, is likely what it's doing. It's like a maximally boring controversial opinion, wake me up when it starts endorsing the bicameral mind or something actually weird.

Maybe it can produce interesting insights. But it's like asking a pundit for a soundbite: what are the odds they aren't just shuffling together something that sounds appropriate to the question?

How sure are you that they're synthesizing an opinion vs just grabbing some existing opinion ("humans value in-group belonging over truth" is not a new idea) and regurgitating it? That they all agree on this kind of suggests as much. Which of course means that they are, ironically... just following the crowd.

It's really easy to come up with an opinion-shaped sentence, that doesn't mean any actual reasoning has gone into it. It's just a pundit-shaped sentence, something someone might say at a party. It's a maximally bland "controversial" opinion about humanity.

It's far more likely that the AI got this "controversial idea" from human-written text saying that other humans are groupthinky, not from observing humans being groupthinky.

Well yeah, if you enjoy doing Activity then the prospect of having to get a new job and therefore spend far less time doing Activity might be understandably upsetting.

If you can't get paid to think about math anymore then you are by necessity going to spend a lot less time doing it.

What about the people who are in it to think deeply about mathematics questions? It's a lot harder to feel like that's worth doing when there's a black box with a crank that takes money in and spits out mathematics answers.

Like calling Lee Sedol a pretentious tool for retiring in the face of AlphaGo. If you really do mathematics for the enjoyment of doing mathematics, following a roughly unbroken line of tradition for at least a couple hundred years, an alien computer mind showing up and smashing everything would perhaps be upsetting, don't you think?

It doesn't really matter for most purposes. Determinism doesn't make them any less chaotic, it's not like fixing the seed makes the output any more predictable. Some percentage of seeds will behave in ways you don't like for a particular input, and you can't know which beforehand.

First, Don't Panic! and consult your copy of Dr. Dan Streetmentioner's Time Traveler's Handbook of 1001 Tense Formations.

You can bootstrap a tiny duration of prescience into arbitrary durations by passing back the same message over and over as many times as you like.

If you know what will happen in one minute, write down the message you see yourself writing down in one minute. In a minute, do the same thing. Now you can pass messages back two minutes.

The GUI parts are or were.

And parts of their patient portal website were still VB6 on the backend a decade ago, not sure that's true anymore. Visual Basic 6 as a web backend.

Some of the delta is newish:

https://openrouter.ai/blog/insights/opus-47-tokenizer-analys...

Right, yes, this sort of style certainly can be produced by writers, especially ones who are working in a particular corporate universe. That's where LLMs got it from! It just sucks that everyone "writes" like that now. Every new one-person project or company is incongruously spouting corporatese. Don't we get enough of that already? Flattening everything into American corpo-speak sucks.

The website is obviously and gratingly LLM-voiced. I don't know how people aren't tired of paragraphs like this yet:

"Visual FoxPro is a 32-bit program, and that decides more than it appears to. It is why a table stops at two gigabytes, why a memo file stops at two gigabytes, and why a big report runs out of memory on a machine with plenty to spare. The limits are signed 32-bit numbers buried in the file handling, not a licensing decision anybody made."

and

"The editor checks what you type through that very compiler, so what it underlines and what the runtime refuses cannot drift apart."

Like, what? Who talks like that?

It's not hard to sandbox Pi without adding anything substantial to the actual size of the system prompt, which is what HarnessTax is mostly measuring. Wrapping it in nono.sh costs approximately zero tokens.

(I do think Pi should ship with more than zero builtin sandboxing though)

I've found the experience of using Pi with local models feels a lot snappier than both OpenCode or Claude Code.

For me, it's simply that I have to read Claudese instrumentally for work enough that I am not going to spend my neurons reading it idly. I am not mad about it or anything, it just sucks to open up a blog post and realize that it's just Claude again under a different name.

> reach for, a Rust extension does the work

ai; dr, sorry

No less than Shazam is?

Pangram flags it. It also talks about "quietly" and "the gap". It's AI.

ai; dr, sorry

If there's a machine that automatically turns electricity into proofs then mathematics becomes a different thing altogether.

Another effect may be that with newer models, people try out the hard problems they got stuck with in older models, and when that model happens to succeed, they are very impressed.

Perhaps the new model really is better... but perhaps merely being prompted to try again with the difficult problem just gave them a new dice roll and they came up lucky. And then regression to the mean kicks in.

Today, maybe. Where's the law of nature that says it won't be generating novel insights in two years? Five? Ten?

If AI can generate questions and then answer them, what are the people for?

> I chained this bug with an n-day sandbox escape and flagged the v8CTF.

https://serotav.github.io/Writeups/v8/when-sorting-leads-to-...

A similar site, also LLM-assisted, that does a better and more evocative job of composing stories.

https://random-lives.github.io/random-lives/

> The subsequent creation of bigger explosions by packing more fissile material does not meaningfully improve the technology.

That's not a great summary of what happened with the invention of the H-bomb.

Fable zaps anything that seems even vaguely biological.