Post

"why only glibly mock those with valid critiques" Glib mockery is the NYT's consistent editorial position on valid critique.

Och you're so close

How are concerns about content rights and environmental impact predicated on it being useful? If anything, I often see the latter framed as "it uses up all these resources and we don't get anything good in return"

There may be some kind of cleavage in life between those with something better to do and those with something to gain.

Along similar lines… bsky.app/profile/neil...

I get this all the time too. I have to correct people and say, "wait, my critique is not based on Ludditism... I cofounded and still codirect @hastac.bsky.social "the world's first academic social network." It's precisely from that base of knowledge and investment that I frame my critique!

Yet.

"Why make the conversation dumber?" Roose first showed up on our radar praising the boldness of the hyperloop, so this is definitely on brand.

The goal is for the Times' designated mark to once again hand over his wallet to the Silicon Valley wallet inspectors. www.nytimes.com/interactive/...

Are you saying that the article talks too much as though there are people who deny that it's useful and/or works, whereas really hardly anyone denies that? (For some values or "useful" and "works"?) I think lots of people do deny that, for SOME values of those words. And that's fine.

Tell that to the people in tech who have lost jobs or paid hours due to employers deciding it already is.

Frankly, his attitude toward this site and his AI-superiority-over-all-tech has caused me to stop listening to Hard Fork and clicking through to read any pieces he has written.

Because Kevin is a troll, and he is not interested in genuine engagement

I think it’s not an equivalent example because Bluesky has sort of become known as *uniquely* anti-AI, due the huggingface thing. It feels more unified than suggesting X has a pro-AI stance because it’s just bigger and so less coherent in all of its community positions

GuYs Im JuSt AsKiNg QuEsTiOnS i WaNt To HeAr BoTh SiDeS wHy Is No OnE TaKiNg Me SeRiOuSlY

As many people have unsuccessfully pointed out, there have been simple websites doing exactly that for decades, no AI needed. And they didn't ever recommend glue on pizza!

Srsly, I'm old enough to have seen microcomputer ads promising this.

or, famously, glue with pizza

The tragedy of genAI is that companies could have directed this insane amount of money and effort toward making useful tools that actually help people but instead they’ve gone all in on a bloated tool that does everything fine to poorly and in the most inefficient way possible.

Social media destroys people’s brains.

Useful for what? My sense of AI as it exists now is that it can be good for doing calculations or tests at volume, or organizing/synthesizing information/data you have specifically given it in a closed system. Its generative work, in my opinion, is horrible. Truly shit.

Seriously? See under historians and academics concerned about learning

Salesforce is selling shovels during the Gold Rush. They're not generating value from the AI itself, but from the myriad of companies who are buying in.

AI is not useful and it’s the most garbage way I’ve seen humans throw away deductive logic/ critical thinking. Outsourcing everything makes us even dumber. Not a soul will be able to stand against fascism when no one can figure out WHY it’s bad anymore. Maybe that’s why the uniparty is pushing it.

you're missing ... data security ... personal privacy ... undercutting education and learning by doing. Let AI have its monetized way and people will forget how to write, speak, think properly.

Generative AI is totally unserious bullshit that often requires hours of undoing its "work" by professionals. If you're like an idiot who doesn't know the difference or you need something super basic and you're willing to do the 50/50 coin flip it won't give you slop, I guess fine, go off

Preemptively muted, so don't fucking bother.

AI is not useful. Only for bullshitting at best

It doesn't have to "work" for it to present those dangers. People only need to FALSELY BELIEVE that it "works" for those dangers to be realized.

"Artificial intelligence" is like calling a cart pulled by a horse an automobile.

It works (as in: it does stuff) and it's useless because the stuff it does is mostly garbage and users don't realize that (and why would they - it *does stuff* after all and people hype the everliving shit out of that!).

You didn’t read the replies, did you?

*AI* is useful. Generative AI is not useful for most of the tasks it is propounded to address.

It's useful primarily for the people who made it and those adjacent to them. That makes them convinced that it's useful for everyone, and it's just not. It's not useful for me in the slightest given what I do for a living and a lot of boosters are unable to leave me alone about that.

from the words of Chuck D Don't believe the hype it's a sequel

It's more work editing bad and inaccurate writing than writing from scratch

Oh I agree. Sadly the mobile web is mostly unusable now without an AI to cut through the ads and promos and special offers and pre roll videos and whatever else that’s there instead of content.

It tells people to eat poisonous plants, dude!

No I think you are more precisely analytical and I was being a bit arch about the category of shit emails.

If you conveniently define all critics of AI who say it isn’t useful as unserious, then I suppose you are trivially correct, but lots of serious people criticize LLMs as useless, over-hyped parlor tricks, even without the legal and environmental issues.

Wait works too well for people to manage? Idgi how

Also: the second part of this is a non sequitur. I am concerned about all the things you list even though LLMs work quite badly. All of those things are happening / will happen *because of people in power who believe it works well* or worse, know it doesn't and don't care.

I see a lot of gaps in AI. I don't see how the AI we have today is going to replace human workers. I don't want to be helped in a customer service sense by an AI chatbot. If AI is replacing people who create knowledge, how will AI stay current and up to date without people making new knowledge.

I should be clear. The counter point was to that there are no critics making that statement.

If the proponents of AI uselessness would stop conflating LLMs with “all machine learning ever” there’d be no problem. Also there are plenty of programmers in this very thread saying “Even LLMs are useful to me” so forgive me if I don’t believe bringing evidence to the “public forum” would help.

AI is useful just like Radiation is useful. Both have good uses and bad. Things like Waymo are useful for disabled folks. AI is sinilarly useful in that it allows things like Netflix and most of the internet to function, e.g. it protects from DDOS attacks. There’s more there than stupid chatbots

I wouldn’t argue that. But there are people who claimed to find NFTs useful.

One problem is that tech firms "selling" AI are pushing generative AI for everything. Another is a knee-jerk reaction conflating every tech firm together.

You make it sound like epistemological threats are a minor side-effect and I dunno, sounds like accelerating the intellectual death of America is bad.

Maybe if the people promoting AI didn’t push a version that sucks down our throat, while whining that they need special treatment people would think it’s useful. They can go back to the lab until they get it right.

I may not be a serious critic, but all the proven use cases of AI are “bulk thesaurus” and “google front end that pretends to speak in first person” granted, there are porno implications.

But GenAI doesn't really have any practical uses beyond being a curious toy. Any applications you could use it for has a specialized tool that is cheaper, more efficient, and better suited. Even if it's other drawbacks (one of which is "it doesn't work most of the time) were fixed.

We don't have to. If LLM proponents want to say the gizmo is useful for a thing, THEY are making the claim. THEY have to provide the evidence. This, I have not seen. The times when people get out and measure results come to the conclusion that LLMs are less useful than a user manual.

I don't use because the times I have used it it's been useless.

God, those typos, whatever my point still comes across.

AI Sucks entirely, what planet are you from. ugly art, incorrect answers, techbro nightmares.

Sorry about that. As a specialist in the economics of information, I spend a lot of time immersed in the “I know that you think that I believe …” of things, and got triggered.😢 I wish the term “AI” didn’t exist.

Generative AI, as it is, is not useful. Iterative AI has useful capabilities in certain fields, but not gAI. I consider myself a pretty serious critic about this.

Well said Anil. As you note, most AI criticism is focused on the exploitation of humans & earth, & the consequences for humanity/society. There’s a separate strain of AI criticism that’s more technical—but it’s largely arguing that LLM’s are not the pathway to AGI, not that they don’t work at all.

This is perfectly stated. Thanks @akalhan.bsky.social

I know you know this but Newton has been setting up that straw man (the skeptic who thinks it’s fake) for a while now. Wrote a little bit about that here. joshuawood.net/llms-debugging

It does not work. It spews out garbage. And as it spews out more garbage and takes in other AI garbage to train on it will get worse.

Absolutely. People also need a critical thinking framework to assess AI uses. We often jump to ethical considerations before even looking at practicality. When people regurgitate tech bro hype, I try to focus on the actual tools and workflows they are using and if they're even useful.

Yeah, I think OpenAI in particular encourages toxic uses of its platform like that. This is part of what I want to pull apart, those spurious, even dangerous, assertions from vendors of what’s possible (or responsible) to do, vs what we could teach current users about to minimize harm and risk.

Depends on your definition of "useful" and of "AI". If you're describing AI broadly, of course it's sometimes useful. Depends on the AI. However, with generative AI, if your job is to make other peoples lives worse, it seems to work great. If you actually want to improve the world, not so much.

📌

Explain how environmental effects are predicated on it being useful

Llms have very niche very narrow uses but the cost is absurdly high for those minimal uses. Imagine a technology that let you get to work 10% faster than your car... But every time you used it it killed 1000 random animals. Would that be "useful"?

It's more that it's being oversold as some kind of cheat code for pretty much everything, when it's actually atool that has some valid uses that will make an impact but won't make investors nearly as fich as they'd like.

I don’t have any of those, and I can do this when my hands are full or dirty.

It's being used by some people. That doesn't prove it's good. There's quite a difference between those things.

I think I agree with you, but there's a connotation of "useful" as "useful for something good", or "useful at what it's supposed to do", which is invoked by people saying, "I'm not saying it's useless...". I am a seriously critical person who would not hand that plaudit to AI (particularly LLMs).

So aside from morality and probity, I fail to see the point of it, except to enable those who won't put in the time to write or make art at a level that satisfies them kid themselves that they can. For this, creators' rights have been stolen.

Yes there are. There are several serious critics who are honest. There is not a use case. Every time they come up with one it almost immediately comes out that it makes too many errors. It's always "someday..." How long does something have to be in public alpha testing before you let it go?

I think any discussions of “seriousness” need to refer to the *type* of AI. Well-trained machine learning models for reading medical imaging are, most certainly, useful and realistically promise more equitable access to “expert” diagnosis. The *generative* stuff (LLMs)? Not so much.

I say this and yet I also think we should keep an eye on if it is sentient or not, to maybe treat it more kindly. zero love for the corporate powers that are pushing this, yes, but like...looking for sapience in unexpected places is a moral obligation I think.

or when it does successfully write rust, it's cloning every two lines, or double boxing data.

Yeah, Ed Zitron is unserious. He doesn't even parrot what tech CEOs tell him uncritically, what a buffoon.

No worries. I know this stuff is so fraught and stressful, especially out of context. Appreciate you being willing to listen.