jprete

I believe this is true, but I'm curious to see an example, if you have one.

I think the problem isn't specific to SO. Text-based communication with strangers lacks two crucial emotional filters. Before speaking, a person anticipates the listener's reaction and adjusts what they say accordingly. After speaking, they pay attention to the listener's reaction to update their understanding for the future.

Without seeing faces, people just don't do this very well.

I think the actual reason people can't do it is that we avoid situations with high risk and no apparent reward. And we aren't sufficiently supportive of other people doing surprising things (so there's no reward for trying). I.e. it's a modern culture problem, not a human brain problem.

Ozempic does not simply limit your appetite, it seems to also affect how much reward your brain feels from different foods (and activities!), which would make it easier to override those anticipated rewards with conscious choices.

I think the problem is inventing new structures for logic games. The shape of the problem ideally would be different than any existing puzzle, and that's hard. If a person can look at it and say "oh, that's just the sheep-wolf-cabbage/liar-and-truthteller/etc. problem with extra features" then it's not an ideal test because it can be pattern-matched.

You don't trust AI-generated content, and your response to this is to make a tool to fool people into not knowing it's AI-generated content?

"About sixty years ago, there were only about seven different web sites, and none of them had any user interaction or even navigation. They just autoplayed the same scheduled streamed video to everyone who looked at them, all the time, like a YouTube autoplay you can't turn off. If you wanted to talk to the YTer, you'd have to do a realtime voice chat, or send them a DM on a piece of paper, and they usually wouldn't even answer. There were also podcasts - more of them than websites - and these were mostly also only sent once and couldn't be listened to again. If you tried to voicechat the podcasts, they would often respond, and might even talk to you live on the podcast. Since there were so few podcasts you'd probably listen to the same ones as your friends, so there was a chance they'd even hear you talking on the podcast."

That guy is more techbro than I can really handle.

Hinton already won a Turing award, so this Nobel just seems doubly absurd.

I hope he turns it down, but it's a monetary prize too and it takes a lot of dedication to science to turn it down.

The decreasing human population is the problem. Rather than let the cats suffer and starve after the last human leaves, they sterilized them so there would at least be less needless suffering.

That's well above the threshold for destroying even the pretense of photographs having any information value on the Internet.

> Although it's conceivable there's a surprise legal finding, companies like OpenAI and Anthropic are confident enough in how it will go that they are willing to insure you for any lawsuits, which would be ruinous for them if they consistently lost.

I don't think this is the right interpretation, at all.

They can act confidently about this because corporations can't go below zero; the downside of the bet is so far in the red, in that zone of bankruptcy, that it actually makes the bet work even if they internally believe that they're likely to lose in court.

Related - thinking takes a lot of energy, so people prefer options that are cheap to evaluate. This definitely contributes to the preference for expedient options.

What does this look like from the photon's frame of reference, then? Or is it nonsensical to try to describe things from a frame of reference at C?

Nobody cares about driving around in a million-dollar car. They want the money/power/status of the person who owns the million-dollar car. An unearned million-dollar car is practically a liability instead of an asset.

Bear in mind these systems have already been through the reward-based training, and these are the results that are good enough to show in public.

I would really like to hear from an actual physicist on this question since both of you seem correct for one of the reference frames and the only way I see to reconcile the two is with weird singularities like "the particle reaches the edge of the visible universe in infinite time according to the rest observer". (And I don't think that is right either.)

And the other half of this is that brains are extremely complicated casual chains. Causality can traverse the diameter of the brain something like 10^8 times in the period it takes for light to be perceived.

I wonder if there's something about tech culture - or tech people - that encourages them to really, really like snowclones.

The upper and lower bunk graphs are symmetrical in their probabilities, so it would intuitively seem like connecting to the same bunk's point would be easier/more likely than connecting to the other bunk's point. The first would require zero (or an even number of) crossings, the latter would require one (or an odd number of) crossings. Every crossing would seem to only decrease the odds of the path existing, or at best leave it the same.

To me it smells like a trap, though. This feels like exactly the kind of problem where some algorithmically chosen infinite graph would break the conjecture, possibly through some kind of NP-hardness-proof-style transformation of the problem into another form, and who knows, maybe the Axiom of Choice gets involved too (but probably not based on the headline!).

The possibility of well-written AI prose is a big reason why I dislike AI text generation. Writing for an audience refines the ideas being communicated. If the author doesn't do that refinement themselves, then I'm not reading what the author is thinking, I'm reading what the LLM could patch together from their ideas. If I want the LLM's opinion on how to make a concept work, I can ask it directly. If I'm reading something by an author, I want to know what the author is actually thinking!

The government at least doesn't have a direct incentive to screw it up and can be pressured by a lot more groups than just industry lobbyists.

I think you misunderstood the OP; they were comparing two different conditions (Java slop vs. LLM slop), because they think there's a common cause (tooling lets people make overly complex systems that still kinda work but are difficult to understand).

As far as I could tell the OP was not saying, at all, that the LLM slop was written in Java.

I'm not a physicist but as soon as I saw it I thought that it would be very costly (in physicist time) to distinguish it from a real physics paper. The difference between this and the output of a physics crank may well just be how fast they can produce it.

This is very 1990s chain-email glurge story.

"The Democrat decided to reject the measure because it applies only to the biggest and most expensive AI models and doesn’t take into account whether they are deployed in high-risk situations, he said in his veto message."

That doesn't mean you're wrong, but it's not what Newsom signaled.

I really wasn't sure!

But to give you a more substantive response, I think the tastemakers are not in control in the slightest. Everything about Notion now screams "we don't give a crap about note-taking or personal wikis, we want a piece of that juicy enterprise market".

Copyrighting, or copy-writing?

It never occurred to me before that chatbot randomness might have the same reward structure as a slot machine, but apparently it does. OpenAI got you to spend all your credits on seven attempts at this one problem. I'm not saying you're addicted, but I wonder about the people who absolutely gush over it.

At this point I'm so used to this that I mentally translate "for health" into "to look good for the opposite sex".