We talk about tribalism and polarization in the modern world a lot, and a lot of the time we correctly focus on the lies and obvious nonsense that people on the right believe, but here's a great illustration of the same thing happening on the left. Read the skeet. Read the replies. - ThreadSky

danielzklein.bsky.social • 57 days ago

We talk about tribalism and polarization in the modern world a lot, and a lot of the time we correctly focus on the lies and obvious nonsense that people on the right believe, but here's a great illustration of the same thing happening on the left. Read the skeet. Read the replies.

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

roguemarvel.bsky.social•56 days ago

I know people leading the legal fight against generative AI and many of them admit to there being decent use cases for LLM and other AI but the lack of ethics and the failure of any type of meaningful oversight and the rush to push ethical bounder to please investors poisoned the nuance.

danielzklein.bsky.social•56 days ago

Yeah absolutely, and I'm entirely on that side. What I'm calling out here is people who are opposed to gen AI / LLMs feeling obligated to say there are absolutely no use cases.

roguemarvel.bsky.social•56 days ago

I think that's why they say there is no cases, because it's gotten so bad and so polarized there is no longer room for nuance in many spaces
I don't see many people who in the beginning were more open to different types and use cases talking out that anymore..it's. All Anti AI

samguyjones.bsky.social•57 days ago

This is probably stronger in your industry than on "the left" overall. I definitely understand anyone in the games industry feeling under attack right now.

And about 95% of the articles talking about AI are describing a bad application or making a bad assumption.

samguyjones.bsky.social•57 days ago

So when people say, "AI makes it so you don't need developers anymore." People say, "that's completely wrong."

My experience is that AI can provide a kind of sloppy junior developer, which is kind of amazing. I can fix that a lot faster than I could write that.

samguyjones.bsky.social•56 days ago

And, while the anti-AI view lacks nuance, it doesn't deserve to be compared to MAGA style tribalism, which has a lot of elements based on fundamentally wrong assumptions.

danielzklein.bsky.social•56 days ago

Well depends on what you mean by comparing, right? Is it as bad as saying vaccines cause autism or climate change is a Chinese hoax? Hell fucking no, not even remotely in the same league. But is it at its core motivated by a similar mechanics (tribal belonging over truth)? Yeah, I'd say so

danielzklein.bsky.social•56 days ago

Yeah that tends to be one of the use cases for me: get me a shitty first version in something I know well enough to fix. The other use case that works really well for me is "something sufficiently complex that it would take a while to learn but that I don't use often enough to justify doing so"

danielzklein.bsky.social•56 days ago

I keep returning to the example of ffmpeg command line options. Because in spite of the fact that I've been computer literate my entire damn life, ran Linux on my home PC for half a decade in my early 20s, and admin webservers for fun, this shit is INDECIPHERABLE to me: https://ffmpeg.org/ffmpeg.html

danielzklein.bsky.social•56 days ago

But when I say to ChatGPT "hey make me an ffmpeg command line that re-encodes this 4k video in 1080p at decent quality, cropping the top and bottom 10% off, and muting the audio" I get something that works instantly. Whereas myself? I would have spent an hour+ trying to get that to work.

nhales80.bsky.social•56 days ago

IMO, it isn't about "can AI save you some time personally", it absolutely can. Is the cost to the environment, society and industry worth "saving you a little time"? I don't think so. I think in certain applications, i.e. medical research, it probably is, but it's all or nothing right now.

lee.scheinbe.im•56 days ago

there are many, many free GUIs that wrap around ffmpeg of varying ease to use. i use handbrake. why is chatgpt better than using one of these?

terrified.bsky.social•56 days ago

I was discussing this with my wife yesterday - laying out all of the reasons why I hate ChatGPT and other LLMs, hallucinations, IP theft, environmental concerns, all sorts of things. But even I concede that she, who is veeeeery dyslexic, could find excellent use cases for it. Are there better…

terrified.bsky.social•56 days ago

…options? Sure. But as you say, there’s better options than Amazon, and we are often forced down that route. Eventually, I suspect there’ll be smaller, almost single use types of these software to cover such uses that don’t set the Amazon on fire, have more ethical training datasets etc.

danielzklein.bsky.social•57 days ago

I think I've been pretty damn vocal about this, but I should say it again here: LLMs and generative AI are fucking awful and should not exist. They do exist because the rich couldn't find anything better to hype up and overinvest into. They're ethically wrong and an environmental disaster.

danielzklein.bsky.social•57 days ago

For almost every use case that gets advertised they're extremely bad, and the people trying to sell you AI don't understand anything that's happening.

But.

There's absolutely use cases where using an LLM is the fastest and easiest way to get to where you're going.

danielzklein.bsky.social•57 days ago

Does that make it ethically right to use them? Hell no. It's also wrong to order from Amazon, use an iPhone, wear clothes manufactured for starvation wages in the global South, and so on. We should do what we can to offset the negative impacts of our consumption and choose better options.

danielzklein.bsky.social•57 days ago

But Hank's not arguing for "everyone should use LLMs all the time" here. He's saying there's actual use cases where they're useful. And I'll add: lying about this isn't a good strategy. But it's not even a strategy here I fear, it's more of a purity test, a ritual of belonging.

danielzklein.bsky.social•57 days ago

We are politically on the left, therefore we must hate LLMs (and again, I do hate that they exist as theft and pollution engines), and therefore we "must" say there's absolutely no use case ever where they'd be useful. And that's just neither true nor helpful.

danielzklein.bsky.social•57 days ago

When you make your arguments based on lies based on emotions and tribal belonging, your arguments are weaker and easier to dismantle. But more importantly: you impair your own ability to think critically.

Comments

Posting Rules

Reply