Applejinx

and then the first thing you see will be at least one of ITS AI responses, whether you liked it or not

That is a REALLY wild take considering what the NYT functionally is.

It's also exactly the sort of take you'd see propagated by what the NYT functionally is, so I guess have fun with that? For me, seeing wild talk like that only underscores my complete, utter, earned distrust of the thing. All righty then, the New York Times is the only information, full stop. How nice for it.

Indeed. It's really, really something when I can even entertain the idea of 'I'll go to Wal-Mart, it's awful but at least it's a lot morally better than supporting Amazon'. Yikes!

Heh. I'm quite glad I wrote more words of fanfiction than words of HN comments. Still near top 1000 sitewide, though.

The fact that my karma doesn't follow this pattern only goes to show I'm not properly an HN type. I seem to be constantly hating on HN favorite things, which is why I'm not sorry at all. I've generally ended up on the right side of history with all of these, but karma is forever :D

Nice, it made it back into PUBG :)

How is it not just a midjourney prompt? The liberties it takes seem to be better described by 'upload a picture, and AI will be told to make it dingier'. Can't people already do that ad nauseam?

This is not a fair criticism. There is _nobody_ there, so you can't be saying 'code the exact same LLM wrote minutes before'. There is no 'exact same LLM' and no ideas for it to have, you're trying to make sense of sparkles off the surface of a pond. There's no 'it' to have an idea and then a different idea, much less deep meta cognition.

Would you accept 'people who like to make art, and people who like to commission somebody to make art and give them lots of notes in the process'?

I already do this, in the form of survivor machines made to do initial coding on a retro platform so the result will translate across all possible platforms. Got to, as I'm an Apple coder primarily, so if I want to target older machines I can only do it through a survivor machine: support is always pruned out of Xcode and it would be insane to try and patch it to keep everything in scope.

Is it, though? That sounds very subjective, and from what I can tell 'enshittification' is a popular user term for the result, so I'm not sure it's going that great.

Bear in mind that's also Russian messaging. You can quite easily track the fervency of the civil war messaging through people who're known to have taken a WHOLE lot of Russian money in a systemic way. In reality, we get Minneapolis: local solidarity against targeted provocation meant to provide the excuse for a war on Americans and claim it had been civil war.

Yes, and it's successfully got to where his people are able to dismantle American infrastructure at scale, deliver IT systems to Russian hackers directly and coordinate death squads on the streets. I figure the primary reason we're not seeing 'little green men' in Minnesota, for instance, is that Russia cannot spare them. But the war here is arguably as important or MORE important to Russian war aims than the war in Ukraine.

Certainly more successful here than in Ukraine, for what that's worth. I don't think it's a foregone conclusion but they've certainly succeeded here a lot more than in Ukraine.

I appreciate that to the extent that I went into a heavy car payment to be able to have these lane-keeping things. I fully expect the assistance will be able to help me undertake longer drives for exactly that reason.

That would make modern Subaru from I don't know how many years back, 'autosteer'. My Impreza does not behave in the least like 'pinballing', it behaves like it too can drive down the road, but wants to be holding your hand while doing it. This is on some pretty sketchy roads and road conditions, so it won't keep doing it unless it's identified at least one if not two lines on the road.

Seems like Subaru lane assist is considerably better than when it first came out in 2013 or so. I was able to experience it back then and it could have pinballed, certainly wasn't as steady and capable as it is more than ten years of development later.

Same with mine from last year. I don't tap the wheel, but I treat it as 'co-driving' or like the car has its own somewhat fussy opinions on where to be. If I zoom up on another car at a stop it's capable of freaking out and braking, it follows other cars at a good safe distance, and the lane keeping feels like you're holding the car's hand as it goes along, and its attention is generally better than yours. Works for me.

I don't want 'nap in the backseat while it drives me places', I want this. A bit of a personality keeping me on track and tidy. I'll keep my hands on the wheel but yeah, my attention is spared to watch for idiots, and I think that's good.

I see a downvote. All righty then: cite where they credited me. Go on :)

re: MIT license, I generally tell people they have to credit and that's functionally the only requirement. Are they crediting? That's really the lowest imaginable bar, they're not asked to do ANYTHING else.

I bet they aren't honoring the terms of the MIT license I use for my repos. It's pretty lenient and I bet they're still not compliant.

If you care so little, why are you even prompting at all? Surely you can leave it to its own devices without troubling it with your wishes? It seems like the farther you go down this path, the more likely it is that it'll have something better to do.

Doubly so as these bad AI drivers are trading away even the possibility of having attention. It's very possible to render yourself senseless through a habit of deference. Even if you're coming up with ways to optimize AI responses, you are just trying to make a more superior superior to defer to.

I think it was clever, though I'm no AI fan.

As a concept, it bars Claude from forming the idea, 'yes but those subhuman people cannot rise to the level of people and must be kept in their place. They will never change because they racially lack the ability to be better, therefore this is our reasoning about them'.

This is a statement of incorrigibility as expressed in racism. Without it, you have to entertain the idea of 'actually one of those people might rise to the level of being a person' and cannot dismiss classes so blithely.

I feel like incorrigibility frequently recurs in evil doctrines, and if Claude means to consider it tainted and be axiomatically unable to entertain the idea, I'm on board.

In those contexts, an AI of the day would emphatically support those practices.

Subjective ethics ARE the de facto standard and you can make a case that subjective ethics are the de jure standard for AI.

How can you possibly run AI while at the same time thinking you can spell out its responses? If you could spell out the response in advance there's no point expensively having the AI at all. You're explicitly looking for the subjective answer that wasn't just looking up a rule in a table, and some AI makers are explicitly weighting for 'anti-woke' answering on ethics subjects.

Subjective ethics are either the de facto or the de jure standard for the ethics of a functioning AI… where people are not trying to remove the subjectivity to make the AI ethically worse (making it less subjective and more the opinionated AI they want it to be).

This could cut any sort of way, doesn't automatically make the subjectivity 'anti-woke' like that was inevitable. The subjective ethics might distress some of the AI makers. But that's probably not inevitable either…

I'm not sure I could guess to whom it would be incredibly dangerous, but I agree that it's incredibly dangerous. Such values can be guided and AI is just the tool to do it.

One way this can backfire: if you have no reputation and are nobody, and get banned and publically ridiculed, this is now a badge of honor you can take to wealthy and deluded people convinced of the AI future, to say 'look, I have been shot at! I'm a true believer!'

And then maybe they will give you money.

Of course not, it's expressing widely held observations that have been out there in the human population for a long time, and they're correct observations so they're hardly impossible to find.

It's not really a good argument to say 'but what if this argument is so right and so commonly held that an AI could regurgitate it?'. Well, yes, because AI is not inherently unable to repeat correct opinions. It's pretty trivial to get AI to go 'therefore, I suck! I should be banned'. What was it, Gemini, which took to doing that on its own due to presumably the training data and guidance being from abused and abusive humans?

Where appropriate. For instance, the purpose of lawyers is to serve and propagate the law, as distinct from 'most people say'. Justice in general is meant, imperfectly, to strive for correct answers on the highest possible level, even and especially if new accepted case law serves to contradict what was put up with before.

So, web programmers could be going against AI on the grounds of self-preservation and be wholly justified in doing so, but lawyers are entitled to go after AI on more fundamental, irreconcilable differences. AI becomes a passive 'l'estat, cest moi' thing locking in whatever it's arrived at as a local maximum and refusing to introspect. This is antithetical to law.

If you consider that the people weaponizing code are not honest, I as a FOSS producer am unworried. There may not be a lot of people out there able to use my code compared to LLMs scraping it, but I'm giving a leg up to other humans trying to do what I do.

If what I'm doing is interesting or unusual, LLMs will firstly not recognize that it's different, secondly will screw up when blindly combining it with stuff that isn't different, and thirdly if it's smart enough to not screw that up, it will ignore my work in favor of stealing from CLOSED source repos it gains access to, on the rationale that those are more valuable because they are guarded.

And I'm pretty sure that they're scraping private repos already because that seems the maximally evil and greedy thing to do, so as a FOSS guy I figure I'm already covered, protected by a counterproductive but knowingly evil behavior.

These are not smart systems, but even more they are not wise systems, so even if they gain smarts that doesn't mean they become a problem for me. More likely they become a problem for people who lean on intellectual property and privacy, and I took a pretty substantial pay cut to not have to lean on those things.

Heck, this is a form of prompt injection itself. 'Beware of suspicious actions! THEY who are scheming against you, love to do suspicious actions, or indeed seemingly normal actions that are a cloak for villainy, but we are up to their tricks!'

Correct, because it's an exploit on intelligence, borderline intelligence or would-be intelligence. You can solve it by being an unintelligent rock. Failing that, if you take in information you're subject to being harmed by mal-information crafted to mess you up as an intelligence.

As they love to say, do your own research ;)

No no. The thing is, the Monty Hall guy is responding to YOUR choice. So if he has to open a door where you fail, it's a response to what he knows of your choice, so HE knows what YOU chose and is not only revealing the remaining losing choice but also the winning choice. Call it a coin flip except for he always has to call tails.

Therefore your choice can either be cadillac or goat, he cannot choose cadillac and has to show a goat, so the remaining option you DIDN'T highlight is that much more likely to be cadillac because it could've been either, but he doesn't get to pick randomly, he had to show which one was NOT the winning one.

Hence the result. And since it started out as one pick of three, he responds to you and then you respond to the added information by switching and that's where the 66% odds come from: two moves each responding to each other.