I don’t think it’s in good faith to pretend the extremely narrow use cases that require an expert to read the results of an LLM and the attempts to push LLMs into wide, popular, and frivolous use are the same. - ThreadSky

trungles.com • 3 days ago

I don’t think it’s in good faith to pretend the extremely narrow use cases that require an expert to read the results of an LLM and the attempts to push LLMs into wide, popular, and frivolous use are the same.

Reposted from Hank Green

There are a lot of critiques of LLMs that I agree with but "they suck and aren't useful" doesn't really hold water.

I understand people not using them because of social, economic, and environmental concerns. And I also understand people using them because they can be very useful.

Thoughts?

Comments

pradaldi.bsky.social•3 days ago

At this point I just assume people using LLMs and thinking they did a good job just don't realised yet they did a mediocre or bad job.

mcbiohazard.bsky.social•3 days ago

I think it's the notion that people want to think that tech is just a tool that isn't imbued with any kind of morality which is technically true the same way a gun is just a tool until I pick it up and shoot you in the face with it.

cluniies.bsky.social•3 days ago

Yeah LLMs for broad use is not ideal. But they can be great for research and study purposes with a closed dataset. My med school made a closed llm for studying material from our curriculum and it helped me test myself on difficult material that wasn’t easy to assess in lecture. I still do not trust-

cluniies.bsky.social•3 days ago

Them as a rule of thumb, and those goofs in Silicon Valley should never have unbottled this genie in this unregulated way. But in the world of research they can cut drudgery and help look for new patterns.

cluniies.bsky.social•3 days ago

I hate that they’ve been so normalised amongst students and my class though. Like I don’t trust anything to an open AI that I don’t have personal mastery over.

mcbiohazard.bsky.social•3 days ago

The way children and many lazy uncritically thinking adults will confidently Dunning-Kruger a topic that they fed into the LLM prompt and get misinformation wholesale hallucinated into existence that they think is fact is truly depressing.

qwilman.bsky.social•3 days ago

the best, most practical uses for LLMs i've ever even heard of are all mitigated by the smallest amount of professional collaboration

upongoblinhill.bsky.social•3 days ago

His ignoring the whole "theft on a scale not seen in human history" is really making me question the plasticity of his intelligence.

trungles.com•3 days ago

This is one of those moments where I’m reminded that like he’s a rich white guy and he’s going to extend a lot of grace to other people in the same social strata, and they love this shit.

breebird33.bsky.social•3 days ago

yeah...one of the reasons I unfollowed and unsubscribed a couple years ago, the depth of their pockets was starting to show and it was curbing what they could say

trungles.com•3 days ago

LOL I also had this experience, but listening to them own up to the Musk of it all was a little funny (and not enough to keep me involved)

breebird33.bsky.social•3 days ago

nods nods

spanishgalleon.bsky.social•3 days ago

Hank green is a left leaning CEO. His business donates to charities. His model is free education on multiple platforms.

He made a hospital & spreads awareness

These things are good.

But he isn't advocating (and won't) for the removal of capitalism. A left leaning capitalist is still a capitalist.

upongoblinhill.bsky.social•3 days ago

Class Traitor Hank Green, everybody!

theodeity.bsky.social•3 days ago

an usually those results are riddled with errors

nicterhorst.bsky.social•3 days ago

That's the thing that all of these "but the can be useful" soft takes share: just a total and utter disregard to the REAL harm it's doing right now. It's like the world's easiest trolley problem, and they keep acting like both sides are equal in favour of this theoretical LLM without problems

theofficeghey.bsky.social•3 days ago

and they're ruining the reputation for any LLM now! I know it can be good! I used to want to study it! But now if anything says AI I immediately cringe. I don't know if it's genAI or not so I'm going to avoid the product! It's ruined AI's whole reputation and I don't trust any of it now

nicterhorst.bsky.social•3 days ago

100%!! Like on the art side there are super useful tools that Photoshop has had for years and years, or that fun lil plugin online where you draw your bg in ms paint and then it like... Photobashes a reference for you

Like there are useful things! And now we have to reject all of it! It sucks!

theofficeghey.bsky.social•3 days ago

now I can't even use those background replacement tools anymore it's like are you actually doing something or just using genAI to construct a whole new background. Guess it's back to manual! this is why we can't have nice things!

ericleitzen.bsky.social•3 days ago

Crackpot centrism

trungles.com•3 days ago

Soft-pedaling LLMs into everyday use for the general public is like saying that everyone should personally own a sixteen wheeler because they’re super cool and useful at carrying things, and don’t we all have things to carry. It’s not a good faith argument. It’s not a nuanced one.

matt.bailey.run•3 days ago

flood the zone, monetize what sticks

strategy brought to you by: the dotcom crash

korybing.bsky.social•3 days ago

“I understand people using them because they’re useful” is a flawed argument from the start, too. They’re not useful! They’re misinformation machines that are powered by environmental collapse. The way we currently use LLMs has absolutely no upsides.

doomsdaydad.bsky.social•3 days ago

they're over-hyped and harmful.

toughtink.bsky.social•2 days ago

it comes off as wanting permission from his very large following to use things he knows is bad for society, culture, and the environment if only he can get someone to admit a particular use case as fine. but like he’s too afraid to just admit that so he’s gotta couch it in faux intellectualism?

mdwallach.bsky.social•3 days ago

This is a great synopsis tbh

vmcombs.bsky.social•3 days ago

I was discussing this with a friend who is all in on AI. He didn’t understand my skepticism.
He was using a proprietary algorithm trained on a research library that his company owns.
I was thinking of ChatGPT, sold to the masses as all knowing and all seeing.

socialjoe.bsky.social•3 days ago

Yep. LLMs sometimes work well enough as a type of expert system (if you ignore their costs). Pretending otherwise is silly.

robotjq.bsky.social•2 days ago

Useful is a bad metric.

Slavery is useful. I would like to think we agree that it’s bad.

scohoe.bsky.social•3 days ago

but hank is summing up the two sides of the consumer sentiment, from either "eh whatever" to "yea I use to text mom" the average consumer just doesn't care about anything at that level.

doomsdaydad.bsky.social•3 days ago

also, can we just start saying, "they're over-hyped and harmful" instead of "they suck and are useless"?

Comments

Posting Rules

Reply