You'll be disappointed. In proofs like this, you don't get a number that has some deep meaning attached to it; you get a number from running horribly-complicated calculations on the best-known horribly-complicated approximation method. There's no beauty here, and 186 has no real relationship with the primes. We'll undoubtedly lower this limit further with better techniques (although 2, the probably-true value, might never be reached).
Also, your "surprise" might be because you don't quite understand the distribution of the primes. If you just model them as random events that occur with probability 1/log(n), which is the distribution we know from the Prime Number Theorem, then a little math shows you would expect infinitely many twin prime pairs. They don't get rare fast enough for close pairs not to keep showing up. (Sadly, this probabilistic argument is NOT a proof.)
SnapDragon
2d ago
And I guess Squirtle is the official Hufflepuff-approved choice?
SnapDragon
2d ago
Astra does sound impressive based on OpenAI's own press releases ... but so did GPT5, which ended up underwhelming a bit, but we'll see. I'm sure it still represents steady progress.
I was not impressed at all with the ARC-AGI-3 benchmark. It didn't really represent a task that humans could do and AI can't; it was built around a funky scoring system that exploited some differences between humans and AI. And the "baseline" was set by the best scorers, not the average human scorers. Lots of very unscientific choices there. So since I thought AIs failing it was meaningless, I also think AIs passing it is meaningless.
I'm not sure I even care that much about when people ring the bell and declare "AGI is here". In my book, it kind of already is - but it doesn't match what it looked like in my mind where AIs can just do human stuff interchangeably. They're superhuman in some ways, subhuman in (a shrinking number of) other ways. You just have to experiment and find the ways that it can work for you.
SnapDragon
2d ago
I mean, I would too. I would prefer to be able to write as much software as I want and be able to run it myself, so that puts me at a lower level of regulation than farm products.
SnapDragon
5d ago
I've always loved Looney Tunes, and I was laughing all the way through the movie. Yes, it's not deep. And yeah, I never expected it to compare to Roger Rabbit, which is one of my favorite movies of all time. But I just had a good time. YMMV, I suppose. Sure, Tweety and Foghorn didn't have real character arcs, but I didn't care.
Mute comedy has been a thing since, well, forever. Not even counting cartoon characters, there was Charlie Chaplin, Harpo Marx, the monster in Young Frankenstein, Marcel Marceau (who, hilariously, has the only speaking line in Mel Brooks' Silent Movie), Teller... among many others. The important thing is to emote well, and I thought they did a fantastic job of conveying Wile E.'s thoughts and reactions. Just like in the cartoons.
And, wait, are you saying that the Peter Lorre character was "vaguely racist"? That's a caricature of a real actor who appeared frequently in Looney Tunes shorts: https://youtube.com/watch?v=zFC01Ipfo3g (He's a Hungarian Jew, fwiw.)
Good point about Bugs though. :) I was actually thinking that even back when they were first hinting at his presence in Albuquerque. Definitely a missed opportunity!
SnapDragon
6d ago
Such as the old Ultima series, which invented an elaborate system of 3 core principles that combine into 8 virtues, and built the entire world and all the plot around it.
SnapDragon
6d ago
Hey, I went to see Coyote vs. Acme today (Great movie! Go see it to give WB the middle finger!), and there was a trailer for a new movie by the folks who made "Coraline". If we can count on anyone to give children nightmares in the modern era, it'll be them...
SnapDragon
15d ago
Not bad! And what's awkward and expensive now may not be in 3 years.
SnapDragon
15d ago
I think they're trying harder than ever, but it just turns out that building general-purpose robots is friggin' hard (sorry, Asimov). Human-level intelligence is a tiny little evolutionary whoopsy that's like a million years old. Language? Maybe 100k years. Programming? Don't make me laugh. But processing sensory data and navigating the real world? Evolution has been selecting for that since the beginning of time. We have, uh, a little catching up to do.
What's funny is that I think LLMs are already at a level where they could handle the high-level organization of a robot's work (albeit perhaps not cost-effectively). e.g. looking at the environment and seeing where the laundry is and where to put it. But they're not very helpful for the dumb low-level motions (what we think of as muscle memory). I believe we're making progress on it, but there hasn't been a giant spike in capability like with LLMs.
SnapDragon
15d ago
Hah, as soon as I read the word "astroturf", that was my thought too. We've had several threads about datacenters now, and there was even a weird one about how taking books apart to scan them is the equivalent of Nazi book-burning. In the thread above, OP has tried very hard to elevate the water "issue" to something real, and it's nice to see that he's failing. There are good anti-AI arguments out there that an educated crowd like us will take seriously. It's weird that there are some visitors who think it's worth their time pushing bad ones.
SnapDragon
15d ago
I have higher expectations of even a first time caller!
I highly doubt he's a first-time caller. He pattern matches to some previous posters who were also wording their posts to pretend that a lot of dumb anti-AI talking points are just obviously true. Someone who throws in asides like "...probably because it's being correctly seen as a existential threat to people's lives and livelihoods" is not above using an alt or two to build consensus.
SnapDragon
16d ago
This isn't a hostile comment; I'm just curious. Your capitalization of "Black" suggests that you're in favour of leftist racial politics. But are you aware that they don't capitalize "White" the same way? Neither makes linguistic sense, since they're adjectives not based on a proper noun.
From what I can tell, modern journos decided to start capitalizing "Black" specifically because it's silly and wrong, as a way to signal tribal affiliation (like putting obvious pronouns in one's bio). So capitalizing "Black" will annoy people on the right, but capitalizing "White" will annoy people on the left. Is it your intention to do both?
SnapDragon
20d ago
Wow, you have a lot of faith in 10% of humanity!
SnapDragon
23d ago
Nah, there really isn't any sort of clean line between Canada and the US. My western Canadian accent is closer to the western US (California/Washington) than to eastern Canada (Ontario). Our government has spent decades tossing out free speech and free association protections (AFAIK we're the only western nation to have frozen the bank accounts of protestors our dear leader doesn't like), but that doesn't affect foreigners.
SnapDragon
29d ago
Huh? How did you and @OliveTapenade both make the same mistake when replying to me? We're not talking about experts running a command economy, which I completely agree has been historically disastrous. The "experts" being quoted aren't government stooges, they're techbros working at the AI companies, who want to be allowed to be capitalists and spend/risk their own money.
The people trying to involve the government - to shut down data center projects - are the other side, the one I'm arguing against.
SnapDragon
1mo ago
That's fair. And the same could be said for AI, of course! There's at least an honest debate to be had there. Not like this highly-motivated special scrutiny of datacenters.
SnapDragon
1mo ago
My guess is that it just takes a while for the paperwork to clear. The length of time an investigation is "open" might be pretty unrelated to how much manpower is actually put into it.
SnapDragon
1mo ago
They're not "appropriating" resources. They're negotiating with authorities to be graciously permitted to pay for them. Water, electricity, chips, land, buildings, employees. None of these are free.
SnapDragon
1mo ago
I mean, good? I do not want members of the public to have veto power over what smart engineers are permitted to develop (as long as they're paying for it, of course). Imagine if we'd held a plebiscite on whether universities should continue spending money on this "Internet" thing in friggin' 1973, 4 years after ARPANET began. Imagine if journalists were publishing articles in 1907 asking for the government to shut down research on these new-fangled planes because they're flimsy and dangerous and nobody they know has seen any benefits from them.
Experts have, well, expertise in the subject. They know the technology has a ridiculous number of use cases. But it takes time for it to percolate through society.
I don't think it's quite as bad as that. I'll use "she" in person for someone who, in my judgement, is clearly committed to being transgender. (And I reserve the right for this to be my judgement call.) It doesn't mean I "believe they're a woman", just that it's low-cost for me to be polite. The Democrats may have gone way, way, way too far on gender ideology (especially the coercive parts), but if some people want to be treated as women, why not do so? In most real-life situations it's not actually a big deal. And you don't need to bring the culture war with you everywhere.
Calling everyone "they" just sets my teeth on edge. It's not a fence-sitting position - it's very much in line with leftist ideology, forcing us all to unlearn that men and women exist.
SnapDragon
1mo ago
This is about as far from a mainstream view as the Democrats' crazy gender ideologies. It is not a message that appeals to the average voter.
SnapDragon
1mo ago
Right, that's what "unfalsifiable" means. And an inability to question whether you might have been wrong, no matter the evidence, does not make you Nate Silver. It makes you one of the ignorant teeming masses.
SnapDragon
1mo ago
This seems to be a very common LW take. As somewhat of a doomer myself, I find myself agreeing. For being intrinsically unfalsifiable, the prediction record of the doomers seems not bad so far.
Well, sure, that's the nice thing about unfalsifiability. I don't think there's any set of evidence short of the actual AI singularity that would cause a doomer to sit back and think "huh, guess I was wrong about AI risk". LLMs understand us well and default to annoyingly friendly? Ignored. An LLM notices it's being tested and tries to look up the answers once? THEY WILL KILL US ALL.
(Note that I'm not referring to the Huggingface incident, which I do think is genuinely concerning, and counts as evidence in favour of the doomer position.)
The model clearly knew that it was not doing what the prompters had wanted it to do. It just did not care, because it was trained to do whatever it took to ace it tasks. This has implications way beyond IT security.
Yeah, there's some egg on my face here, because just a couple of weeks ago I posted this:
LLMs don't just understand strawberries, they understand us. If we ask an LLM for a paperclip factory, it's well aware that we don't want it to tile the universe with paperclips.
I don't know the exact ExploitGym setup and prompt, but I have read an account that the test involved being given an exploit and told to make use of it, and that solutions not using the exploit would not be considered valid. In that scenario, it does seem reasonable that the LLM should be aware that hacking an outside company was not intended. So this is evidence that maybe I'm too optimistic when I say that LLMs will not interpret our instructions in disastrous ways.
There's still a massive difference between hacking a company's servers and "kill all humans", mind you, which should not be ignored. It is certainly conceivable that LLMs, with their fuzzy stochastic intelligence, could misbehave in small ways but not large ones. Still, I do need to admit that my post didn't age well.
A good summary (and de-escalation), thanks, and it does fit with a lot of what I recall (except for the family-friendly part - they released some mature games, too). Progress really was crazy for a time, with companies leapfrogging each other left and right. I guess what really started this argument was when I said that "Sierra was one of the companies pushing graphical standards in games, from EGA to VGA to FMV cutscenes", which I should probably walk back a bit, because @SkoomaDentist is technically correct that they weren't the first to release VGA or FMV adventure games. But they released some of the best-looking ones of these eras (some of the standouts I remember include KQ1, KQ4, SQ4, Willy Beamish, Gabriel Knight, KQ6, and Phantasmagoria). Their artists were anything but incompetent.
No, I think that'd leave a lasting mark.
More options
Context Copy link