Most AI use is nonconsensual in some way. It’s used on people that don’t agree to it. People are forced to interact with the output of AI. People are forced to use it.

It is a fun toy we shouldn’t be basing the entire economy on a cult worshiping it

There are neural nets that have been in use in industrial settings since the 70s and they’re obviously not what we’re talking about and they’re fine.

There are neural nets that are being used to try to detect cancer and whatever, and they’re obviously not what we’re talking about, and they’re… highly overrated actually, very prone to replication problems when you look at long term data, but worthy of continued study.

And there’s this one-size-fits-all mostly LLM-based mostly Sillicon Valley based new wave of AI, which is what the whole fucking world is talking about, and it’s not fine. It’s not fine at all.

Obviously the way these things were created is unethical, but unless you’ve got a lot more power to do something about it than I think, it does not matter and I don’t care.

The part I want to talk about is whether they’re useful, which is fraught, and I want to break that apart into some points:

• The vast majority of things AI is being used for is either bad, or do not work. Regardless of what it is capable of, the reality of use is that is used for scams and spams.

• Skill atrophy from using AI has already been measured.

• Remember when every fucking person was talking about cryptocurrency? Remember how a substantial portion of people who should’ve known better were trying so very hard to come up with a way the technology could be used for good? Imagine how many more people are getting duped like that now, with a technology that does, in fact, actually do something.

• Every headline about AI achieving whatever mathematical or coding milestone may be true, or may be false, or may be overblown, but it was always always always produced by a marketing department, for marketing purposes.

• People keep proposing workflows where your roles is to constantly check on the LLMs outputs. This has never worked in any other context. You will experience decision fatigue and your standards will decay until a constant percent of the LLM’s outputs get through, regardless of whether it’s on a hot streak or a shit streak. Every AI-based programming flow amounts to this, which means every AI programming flow becomes vibe coding when given enough time.

• Measuring programmer productivity by how many lines of code they write was already a common joke on the internet before many of you were born.

• Gee it sure is funny how these LLMs are so good at topics you know jack shit about, and make disqualifying errors every two sentences in topics you have expert knowledge in.

Despite all of the above, the technology does actually do something. It is not crypto again. But its uses and impacts cannot be understood during the marketing hype bubble. People are not capable of rationally evaluating something that’s hyped this much. And the truth will take way more time to figure out than the time it takes to get to the next part of the AI hype train anyway.

Despite the hype bubble, there is not actually a hurry to figure any of this out. Maybe it’s the real deal, maybe you’re missing out on the thing that will revolutionize everything. Maybe it’s not there now, but it gets there next Thursday. So what? Do you lose anything by getting around to it next year? The time pressure is coming from OpenAI’s marketing department. And yeah they need you to get on the AI train soon so they can make money, but you don’t owe Sam Altman shit.

My understanding is that there are absolutely technologies that fall under the umbrella of AI that are useful and good and should be embraced as long as they don’t come from a fundamentally anti-human place… But none of those technologies have outputs that will wind up here. Basically anything AI that makes its way to hexbear is likely to be a thing we shouldn’t tolerate.

I think that not being 100% anti “ai” should be a ban worthy offense.

I don’t give a shit that China makes llms too. You don’t have to buy the CIA lies on China to recognize they do bad shit sometimes. It’s bad that they sell the Saudis weapons that get used on Yemenis, it’s bad that they make llms, etc. “China does it” isn’t a good excuse for a technology that is anti human in every conceivable way.

It’s bad. I am fundamentally against it and it shouldnt exist.

it’s a technology that while useful in principle was invented for oppression’s sake and will only be applied in the most anti-human way

I have no desire to weigh in on the conclusion of this discussion, but in this part you’re just describing any sort of economically productive technology invented at the behest of capitalists and many technologies that aren’t economically productive or that were invented in other circumstances. That’s just what the mode of production does.

(I’m broadly anti-AI, I just don’t think it’s that interesting to talk about and have nothing unique or insightful to say about it)

It should be whatever I am thinning at the time, of course

CHINA

It has vague uses, but those uses are so limited and niche, and its drawbacks and its potential to mislead people is so enormous, that I honestly think people on average objectively live better lives without it, no matter its method of employment. The evidence that it degrades your ability to critically think is pretty solid.

Anecdotally, AI is insidious. It overinflates peoples’ sense of their own abilities. And its whole deal is saying things that sound correct without necessarily being correct. So unless you’re very confident about what is ‘correct’, and willingly to close scrutinise every last word, and every last amibiguous wording, you will really frequently misled about reality by AI. So much of the ‘facts’ it says are not real, but people believe it all too willingly, because that’s human nature. Personally, I think people who frequently use AI are just increasingly frequently wrong about… everything.

This isn’t me saying “I’m super smart so it’s okay for me”. But it’s like saying everyone should have plutonium. Yeah it could mean we have cheap electricity, but regardless of its ownership, of the way it’s employed, etc. aside from a few people who handle it super carefully, nobody will net benefit and then we all die of radiation poisoning before we realise it.

The majority of people rightfully don’t interact with plutonium, not because it’s useless, nor TOO POWERFUL, but because it’s harmful in invisible and hard to quantify ways - should be the same attitude for AI.

I implore every person, and every socialist doublyso, to be extremely careful and skeptical in every instance where you consult, rely on, or believe AI.

DESTROY THE THINKING MACHINES!!!

The NSA has a warehouse full of programmers that do nothing but search for vulnerabilities. So only the NSA ended up with lists of exploits they can use to make things like stuxnet. They could waltz into any server they wanted too.

But now AI found and closed all those bugs, so the NSA can’t do whatever they want anymore. They’re trying to build a smarter dragnet with LLMs but that seems like a sorry consolation prize since they used to be able to pwn any and everyone.

I don’t use it I don’t like it you can’t convince me otherwise. Small local models fine, if you can excuse the resource burning destructive training, and admittedly it’s mildly better in China than other countries. But I still think they have a stultifying and homogenising effect on creative and productive processes with the reversion to the mean and in the long run will result in many individuals’ basic competencies atrophying as they offload them to the machine.

The site consensus should be the same as my opinion.

I think every technology advance since and including the smelting of bronze was a mistake but I think AI is here to stay so the real question is how to ensure the profits of automation are used for the common good and not just to build yachts and sex clubs for 10 ugly white dudes.

Its ok for me to use. Its bad when you use it.