Hacker News Reader

the belief that AI's are * hallucinating -all while they become exponentially more powerful- is a polite fiction we will use as an excuse to accept the complete domination of reality by these things

i am confused by this sentence. are AIs not hallucinating? is that a fictional claim? am i misunderstanding or is there a "not" missing at the *?

No, I think he's saying that when AIs are better at us at everything we'll deal with it by saying but they hallucinate.

but they are hallucinating. according to this reading they won't in the future but we will pretend that they still do because that allows us to let AI dominate.

that makes no sense to me.

it's ok for AI to run the world because they just make up stuff anyways? don't we want the opposite? in order to allow AI to run the world, shouldn't we believe that they are foolproof and are always able to figure out the correct answer?

https://en.m.wikipedia.org/wiki/Decidability_(logic)

It's kind of a moot point anyways, because Humans are highly averse to making sure their actions are logically sound. Take politics for example, not just the silly day today goings on of it, but more importantly the system that underlies it: it is beyond consideration.

https://en.m.wikipedia.org/wiki/Third_rail_(politics)

"The first rule of Fight Club is: you do not talk about Fight Club."

Humans are highly averse to making sure their actions are logically sound

yes, but the illogical decisions of humans can be intuited or made to be understood, and thus also somewhat predicted. for example, a human deciding over someones social security benefits can be expected make a somewhat rational even if illogical decision, and in particular, past performance of a human can predict future performance, whereas for an AI this is completely unpredictable and therefore has a much higher risk of going wrong unexpectedly. also, humans can be reasoned with, again with a somewhat predictable outcome.

Have you considered the indirection in play, adequately?

probably not. please elaborate

Two forms: culture, semiotics

Missing counteracting force: set theory

i am afraid i don't understand what you are trying to say.

how do these things relate to our discussion?

Culture shapes how one thinks.

Insufficient depth in semiotics shapes how one thinks.

Importantly: it is generally not possible to take such ideas seriously, because of culture. So humanity stumbles half blindly but confidently into the future, oblivious to much of what is going on.

I think he's saying "the complete domination of reality by these things" is going to happen whether we want it or not.

It is kinda interesting. I talked with a less technical member of my extended family over the holidays. Fairly successful guy in his chosen profession ( accounting ). To say he was skeptical is an understatement and he is typically the most pro-corporate shill you can find for a company to save a few bucks. I assumed he would be attempting to extol its virtues with the assumption that lower level work has errors anyway. I was wrong. Sadly, we didn't get to continue down that line since my kid started crying at that moment.

Yeah I'm interested in how it will play out. I can understand skepticism because the current AI isn't that good, but it'll keep improving.

count me among the skeptics. the big problem i see is that there is no way to verify whether any AI output is correct. it is already very hard to prove that a program is correct. proving that for AI is several levels more difficult, and even if it were possible the cost would be so high to make it not worth it.

I am personally somewhere in between. Language models do allow me to do things I wouldn't have patience to do otherwise ( yesterday chatgpt was actually helpful with hunting down a bug it generated:P ). I think there is some real value here, but I do worry it will not be captured properly.

> the belief that AI's are hallucinating -all while they become exponentially more powerful- is a polite fiction

If I may apply a regex:

s/polite/lucrative/g

> at what point does a human not realize or care that he is being actuated by a computer?

I don't know about caring, but I think that the point of the Turing test is to determine at which point a human can't tell whether it's another human. Also, I've read that it's not a particularly good test, because even pre-LLM you could craft an irritating, misspelling, troll of a chat bot and people would think it was a real teenage edgelord.

I think the chinese room thought experiment is another stronger exploration of whether or not computers are conscious (and, I believe, is no longer sufficient). I think people into AI should at least be familiar with its implications (or lack thereof) about human cognition.

https://en.wikipedia.org/wiki/Chinese_room

It is clear to me that humans “hallucinate” all the time, and I don’t see why this should disqualify AI. One prominent human hallucinated that a hurricane was going to go in a particular direction and kindly updated a map, provided by scientists, with a sharpie.

There's a difference between that and asking for a basic fact and getting errors.

Google's AI result, when I ask for the spot price of silver, returns the amount in British pounds, but with an American dollar sign in front of it.

That's not a lie, it's just an absolute misinterpretation.

I’m getting an accurate, up to date price when I perform the search on Google. Can you post a screenshot of your results?

Aaand you’re saying humans don’t do that? There are plenty of humans who will just make shit up with whatever information they have. There’s plenty of humans who wouldn’t even know the British don’t use dollars and would just put a dollar sign there.

I’ve seen a senior engineer at a FAANG confidently interrupt another engineer, who was giving the correct answer, to provide an incorrect answer. They mixed up facts and confidently stated a conclusion that sounded good, but was utterly wrong.