I was reminded more of the alien AI from Constellation Games, which spawned sub sub sub agents to interview humans.
The protagonist sends a message to the aliens asking to be allowed to review the alien civilization’s computer games. An AI submind called Smoke-Cursive-Cytoplasm-Snakebite-Singsong-Polychromatic-Musteline is given the task of contacting him by IM to begin the conversation. Its job is only to verify that they are talking to the right human (since not every human has a unique name) so it is only a simple chatbot and can only understand YES and NO responses. It asks if the protagonist understands and gets a sarcastic NO. It has to contact its parent mind Smoke-Cursive-Cytoplasm-Snakebite-Singsong-Polychromatic to ask what to do next. After working his way up the tree of subminds by answering questions of increasing complexity asked by subminds of increasing capability, the protagonist briefly talks to Smoke-Cursive-Cytoplasm-Snakebite which sets him a task to prove that it’ll be worth an (alien) anthropologist’s time to talk to him.
Smoke-ccs-762d: Well, if it isn’t Mr. Sarcasm
ABlum: YES
Smoke-ccs-762d: Don’t quit your day job.
Smoke-ccs-762d: I’m Smoke-Cursive-Cytoplasm-Snakebite.
Smoke-ccs-762d: Let’s get down to business.
I need to re-read Constellation Games soon. Lately I keep coming back to considering the expectations that its alien society has for the caretakers of artificial intelligence.
Spinning up AI isn't hard for them from a tech standpoint, but since the AI is advanced enough to be considered life, anyone who creates it needs to be responsible enough to be qualified to adopt.
> Giving the absolute maximum benefit of the doubt I understand that they see themselves as "stewards" for lack of a better word.
Only in the same sense that Standard Oil considered themselves the stewards of petroleum. There's benefit of the doubt and then there's just fanfiction. Do not forget that this most aggressive "guardrail" of theirs was not for any safety reason, but just to stop other labs from catching up to their product. They care less about hindering bioweapons, malware, and hate speech than they do free market competition.
this reads like "throw everything at the wall and see what sticks" reactionary-ism... i'm guessing that it's not particularly easy to use claude to help you make bioweapons, and we all know that they have neutered Fable vis à vis security research because people have already been complaining about it. and the funny thing about hate speech is that there is absolutely no need for ai-- it tends to come out the best when spoken directly "from the heart", as it were, anyway.
> Superintelligent AI is more dangerous than a bioweapon.
No, it's not because it doesn't exist (yet) and its further from reach than the other examples. Also, the guardrails are also framed as restricting usage for the development of "competing" products/services.
mixing up bioweapons, malware, with hate speech (which is basically a censorship) shows how very basic people like Trump can win. Hopefully you won't wait to be censored before realizing that anything could be interpreted as "hate speech".
> “Visible safeguards can be probed, so they have to be robust, which takes time to get right,” Anthropic wrote.
Even on Fable, I'm finding that safeguards can quite easily be surmounted just by incrementally escalating the requests. It's harder than ever to one-shot jailbreaks, but incrementalism still feels like a glaring enough issue to make guardrails just a fig leaf of plausible deniability to the media that they care about "safety."
To the average person it means: No matter what task you apply the tool to, that you will be right 85% of the time, and 85% is a solid B, a passing grade, so let's use it.
So in other words, if you show it 100 random people and ask who did it, it will tell you to arrest 15 people, with a 15% chance not a single one of them is guilty?
(assuming it will operate independently for every individual you show it)
That's great.
Of course, putting such a system on public cameras, scanning large amounts of people, is about as stupid as some of the biggest disasters in medicine (using a 98% effective drug on random people, regardless of whether it would help them or not. You REALLY want to check what happens if you're wrong before applying it to a large number of people)
> I don't want to make excuses for AI identification, but no identification process is perfect, it should not be possible that it takes months to clear up.
Indeed you shouldn't make excuses. "{Sketchy component} is just one part of the process and is harmless in principle because we have other safeguards such as... nothing we care to subject to your scrutiny" is the prototypical excuse of a broken system:
> The office stated, “Facial recognition technology is used as one tool among many available to investigators. In this case, it was one tool, but certainly not the only tool, which lent to the probable cause determination that Mr. Richardson was the perpetrator of these crimes.”
The other tool appears to have been good ol' fashioned racism:
> Richardson alleged racial profiling played a role in his misidentification. “I want to say racial profiling. The guy said it was a guy with dreads and a big nose, and then they picked me out of a lineup of guys that look nothing like me,” Richardson said.
It's obvious what any sane society should do in this case, what actual safeguards would be. A sane society would have a social safety net so that being jailed for 3 months and subsequently released innocent wouldn't ruin your life. Not only did he get punished by having to spend 3 months in jail, he also now has to go and find housing, a job, and go through the civil court system, which is even slower, to ... still be made less than whole. I won't be surprised that the police argue and win with a qualified immunity defense.
To make matters worse, mugshots get people prejudiced from jobs regardless of an HNers ability to discern between a charge and a conviction.
True criminal justice, true innocence until proven guilty would have had his obligations to pay rent/mortgage/bills paused, his employer barred from firing him for missed work, and so on.
(I had to keep editing my post - I just want to say I think it's ridiculous that this dude had to be in jail FOR 3 MONTHS)
I don't understand why you would be downvoted. Is your comment raising a pitchfork? Yes. But sometimes when a person's life gets ruined, pitchforks deserve to come out.
> Richardson’s attorney showed time sheets proving he was at work 400 miles away from Florida when the stolen car was sold. Richardson said he has never been to Florida, and his attorney tried to present this evidence for months.
> Richardson alleged racial profiling played a role in his misidentification. “I want to say racial profiling. The guy said it was a guy with dreads and a big nose, and then they picked me out of a lineup of guys that look nothing like me,” Richardson said.
> While he was incarcerated [for two months], Richardson lost his job and his home. He also said he lost custody of two of his children.
Everyone: It's okay to get angry at injustice. Indeed it is the more noble reaction than to shrug and say, "Now let's be reasonable, I'm sure the institution that caused this will redress this."
Those people are people with high HN karma, a lot of it for posting or submitting technical discussions. Is that enough to give that much weight to their opinion on political topics?
>Those people are people with high HN karma, a lot of it for posting or submitting technical discussions.
Click on a few of those names and see what kind of topics they wade into. I wish you were right.
You're not gonna get high karma by having high quality technical discussion. First off no human can likely know enough depth on enough topics to contribute to very many of them. Second off, such commentary only gets praise from the narrow band of people who are capable of assessing it.
Like all "scored" social-ish media this stuff is a numbers game. The way you "win" as these people have is by gaming the scoring system. Post stuff that appeals to the demographics of the site and with a low common denominator so anyone can "approve" of it. And if you click on through you'll find that's what they do.
>Is that enough to give that much weight to their opinion on political topics?
I assure you that a great many of them take great offense when you suggest that they are not experts in whatever they are speaking at a given minute. My personal favorites are the one that appeals to authority by cherry picking links (thereby moving the discussion from one of attacking his opinion to one of attacking his sources, a tried and true troll/propagandist/etc tactic) when you disagree with him and the one tries to portray being involved in politics as though it confers legitimacy.
That said, some of the power users around here are in fact reasonable and seemingly free of obvious ill will.
> I assure you that a great many of them take great offense when you suggest that they are not experts in whatever they are speaking at a given minute.
That I can believe without question. I see it in action every day. But how are they any different from any social media influencer who has an opinion on everything and millions of followers and who use their existing clout to demand more?
One of the people high on that list once introduced themselves by the karma points from HN. That was the value they saw in themselves, that’s all that matters.
That HN leadership list has in the same page the typical serial posters or panderers, with an opinion on every mundanity under the sun, having 10-20 times the karma of other people in the list who bring genuinely unique and fascinating views of things that normal people never get to see or even know they exist.
> I find it pretty irritating to go back to Python (my long-ago favorite) but many people are in the exact opposite frame of mind.
As someone who works exclusively in typed languages for formal methods, what is it you find lacking about modern Python + PyLance? IMO there's still a tiny verbosity issue, and there's no real replacement for fancier polymorphism or (G)ADTs, but I'm very satisfied with it for most things. In particular, null checks are trivial.
It has been about 10 years since Python was a daily driver for me and at that time I wrote it the old fashioned way with no type hints and no static checking, just like grandma used to make. The times I have needed to dig back into it have involved working on old code, so I haven't kept up with modern tooling.
However, in principle any dynamically typed language can be tolerable to me if it can be turned into a statically typed language ;)
But I think I'd still prefer the ergonomics of a language designed that way from the start vs having bolt-ons. My favorite language for the past several years has been F# and I think ML-family languages in general strike a great balance of being able to write terse code when you want to, and being able to model a domain really well with types when you want to.
reply