Live data from Hacker News

The turbulent AI era is here

gatesnotes.com

121–130 of 267 posts

Re: The turbulent AI era is here

#121
post #96
post #86

Earlier quoted context omitted.

They did in fact put radioactive elements in everyday objects back when it was first discovered. Wikipedia has a whole page of "radioactive quackery." https://en.wikipedia.org/wiki/Radioactive_quackery While nuclear power is not intrinsically anti human you can't really separate the public perception of nuclear power and that of nuclear weapons. Anti nuclear sentiment is a lot easier to understand when you realize ev…

None of that is comparable to us having ai buttom on whatsapp, camera, email, jira, vacuum cleaner, everywhere. To workers being threatened with firing if they dont burn enough tokens.

Indeed, comparing AI button in some app to putting deadly substances in basically everything is really dumb. They are not comparable at all and one is much worse than the other.

Re: The turbulent AI era is here

#122
post #14

I've tried to type this up a couple times, each time backing out because at the end of the day, I really don't give a rats fart about anything Gates has to say. I just wish he would disappear.

I mean, I'm as opensource as they come: not used windows since '98, despise microsoft and mostly everything he's done in business. But lets be fair, he moved the needle forwards, and after he retired he's done some fairly good shit around malaria and so on. I don't really begrudge him, if it wasn't him, it would have been some other wanker. We are where we are, better to just let go of the hate.

It's a fairly concise description of the challenges ahead, and hopefully some people read it and think a bit. He didn't have to write it, I'm glad he did.

Re: The turbulent AI era is here

#123
post #102

I feel this is merely stating the obvious. It goes back to the industrial revolution and every revolution before and after. There was a point where they thought a majority of people would be working as switchboard operators and then telephone switching came along. Horses to cars, etc. Jobs transition, new space is created, people are caught in tbe crossfire of supply and demand curves. Incan agree the bullwhip of thi…

> When the data fed into an AI model is based on real data you can get predictability. When the AI inputs start coming from the AI outputs the wobble is introduced which results inevitably into delirium. This is asserted without evidence. There is absolutely no proof that AI requires humans in the loop to function or improve itself. Compare it with Deepmind's go-playing program, AlphaGo, which mainly involved trainin…

Both programs had programmed in to their routines the rules of go (and for Alpha Zero the rules of chess and shogi). The programs then followed an optimization program. Absent the extremely specific "rules of this game" set by a human none of the programs work.

Amusingly, all three games are in the game theory sense trivial because they provably have a correct solution (probably a draw for chess), but the search space is so large that it is unlikely we will ever collectively invest the resources to discover the answer.

Re: The turbulent AI era is here

#124
post #61

Does anyone else feel incredibly nostalgic for the time when Bill Gates and Warren Buffet were the richest people on earth? I remember feeling pretty good about the fact that they were trying to do so much good, and about the giving pledge. Things feel so different now.

Few people were more vicious than Gates when he was running Microsoft. Maybe Larry Ellison but Gates was smarter and more cunning. I don't feel nostalgic about those days and Gates is just trying to buy his way into heaven and a positive view on his wikipedia page.

> Gates is just trying to buy his way into heaven and a positive view on his wikipedia page.

That's gonna be really difficult as the heaven's manager probably have full access to the Epstein files.

Re: The turbulent AI era is here

#125
post #95

Earlier quoted context omitted.

I’m not sure what to say other than you’re entirely incorrect about all of that In fact you could probably make a strong case that the US and USSR did NuclearMaxxx It’s called the nuclear arms race/cold war and was the defining characteristic of society post WWII until 9/11

Nuclear arms race and tokenmaxxing were two differrent things. Not even nearly the same.

Indeed, nuclear arms race was way more wasteful and negative

Re: The turbulent AI era is here

#126
post #102

Earlier quoted context omitted.

> When the data fed into an AI model is based on real data you can get predictability. When the AI inputs start coming from the AI outputs the wobble is introduced which results inevitably into delirium. This is asserted without evidence. There is absolutely no proof that AI requires humans in the loop to function or improve itself. Compare it with Deepmind's go-playing program, AlphaGo, which mainly involved trainin…

>> When the AI inputs start coming from the AI outputs the wobble is introduced which results inevitably into delirium. > This is asserted without evidence. There is absolutely no proof that AI requires humans in the loop to function or improve itself. Nope, it's actually fairly well established. https://pmc.ncbi.nlm.nih.gov/articles/PMC11269175/

That paper has had a pretty turbulent reception and looks pretty conclusively wrong at this point.

It used an incorrect theoretical framing that assumed that data was being replaced rather than accumulated as a result of more training (see https://arxiv.org/abs/2404.01413 which explores this). This is incorrect because this simply isn't how real-world datasets are created via synthetic data generation (which generally accumulate more data over time rather than replace their data). As a result most of the theoretical results were invalid.

Empirical evidence has also cast a considerable amount of doubt on the paper. For example, Microsoft Phi-4 was an empirical test in specifically what happens if the majority of your training data is synthetic rather than human and it turns out that Phi-4 did significantly better than previous models which relied primarily on human data.

There's some nuance to all of this in how exactly you do this, but the original claims of the paper are looking really shaky at this point.

Re: The turbulent AI era is here

#127
post #110

I feel this is merely stating the obvious. It goes back to the industrial revolution and every revolution before and after. There was a point where they thought a majority of people would be working as switchboard operators and then telephone switching came along. Horses to cars, etc. Jobs transition, new space is created, people are caught in tbe crossfire of supply and demand curves. Incan agree the bullwhip of thi…

> new space is created That seems to be the crux here. You think it will be, I (and a lot of other people) aren't sure it will. If new space for jobs are created, I am certain we'll be fine long term. What do you think will happen if it isn't (I know you have strong evidence for and reasons to believe it will be, but just indulge the hypothetical)?

Yes that’s the crux of the risk - will new space be created

On one hand. We’ve never had a technological revolution that didn’t yield more jobs

On the other hand, Terminator

Re: The turbulent AI era is here

#129
post #102

Earlier quoted context omitted.

> When the data fed into an AI model is based on real data you can get predictability. When the AI inputs start coming from the AI outputs the wobble is introduced which results inevitably into delirium. This is asserted without evidence. There is absolutely no proof that AI requires humans in the loop to function or improve itself. Compare it with Deepmind's go-playing program, AlphaGo, which mainly involved trainin…

>> When the AI inputs start coming from the AI outputs the wobble is introduced which results inevitably into delirium. > This is asserted without evidence. There is absolutely no proof that AI requires humans in the loop to function or improve itself. Nope, it's actually fairly well established. https://pmc.ncbi.nlm.nih.gov/articles/PMC11269175/

The key word in that paper is “indiscriminate”, as in:

> We find that _indiscriminate_ use of model-generated content in training causes irreversible defects in the resulting models

If you view AI training as lossy compression of their training data, then lossily compressing the same data repeatedly will result in data degredation; this is well known from other domains (try repeatedly compressing a JPEG image, for example).

That means it's extremely important that there is some content curation in the loop. But there is no reason to believe this content curation must be done by humans, or that it must exclude all AI-generated content by default.

For example, the recent LLM-generated disproof of the Jacobian conjecture would probably be beneficial to include in the training data, despite being the result of an LLM.

Re: The turbulent AI era is here

#130
post #113
post #97

Earlier quoted context omitted.

I’m very familiar with both technologies, but it’s not what my comment was about. The crypto craze would have been the same if we would have ended with some actual use cases for cryptocurrencies. The level of frenzy was completely out of proportion, very aggressive, with that same attitude of „we will impose something on you, change the world, you will be left behind and stay poor forever for having not been a believ…

The thing is - with AI the position that everything will change is reasonable, with crypto not.

Everything is changing all the time. The position that everything will change in favor of the AI overlords is an imposed narrative that has a bigger chance of materializing if they get all people to believe it is inevitable. It is not.
Post reply on HN