PigLick on 28/4/2025 at 04:26
It's worse than you think. I just read the book "Supremacy" by Parmy Olson, scary stuff.
heywood on 28/4/2025 at 13:24
Am I the only one who thinks the hype over generative AI is overblown?
The printing press was truly revolutionary. One of the great things to happen in the US was the growth of public libraries. By the late 19th century most people here had access to countless literary works catering to all tastes and could consume as much as they had time for. In 20th century we got hooked on radio, cinema, television, video games, online services, and the internet. Yet the biggest effect on how we live and how we spend our leisure time came from transportation, not media.
I'm a member of the MTV generation and I remember how cable TV was supposedly going to turn us into zombies. A decade later Springsteen was singing "Fifty-seven channels and nothin' on". We now have on-demand access to anything we might want to watch on devices we can take anywhere, but the average amount of time we spend consuming media every day hasn't changed much since I was a kid. Creating the content on demand vs. selecting it from libraries is a logical next step, but it will become apparent soon enough that watching a sit com is still watching a sit com, regardless of whether you're selecting from hundreds of pre-made ones or ordering up your own.
My kids' generation would rather watch videos of other people doing stuff and share videos of themselves doing stuff than watch shows. The biggest TV audiences are for live events. When we go to movies, we usually prefer to go with someone. Then we want to talk about them afterward. We're social animals and shared experiences are going to be more important to us than highly personalized content.
demagogue on 28/4/2025 at 14:58
I'm thinking at least half of the content you see on social media these days is AI generated.
Some of them are of consequential content. The one I saw a few days ago was a 12 year old girl with her arms around Trump and him leering or something like that, and some expected outrage, which I thought was counterproductive. There are enough real images of him being sleazy around teenage girls, you don't have to generate them.
It made me think about how lazy AI is turning online culture, where they can't even take the 20 seconds to look up an image. They'd rather just generate the same thing. A similar thing you see it is when an AI generated image, otherwise impeccable, has some obvious mistake, like a nonsense word off in some corner. If that were me, I'd take the 40 seconds to open it up on Photoshop, copy a rectangle of the background, and then paste it over the flaw so it disappeared. These kids don't even bother erasing the mistakes. (Well there's an AI option to do that too, but that takes more work.)
Also I still see some people going on long commentaries about their chats with chat bots. I don't know. Sure you can chat in ever longer sensical runs with chat bots, but ... you're not making friends with the chat bot. It's not going to remember you fondly. I wish they'd put that energy into commentaries about chats with other humans that will, and that are actually inspired.
I used to study philosophy of mind. That was my undergrad scene. So I've already got an axe to grind here, that I would like AI to go in the direction of embedding itself in real world experience, and grounding its chats on that. That I think really would be something meaningful. I actually think some of it might be getting baked into the model's gradients now but, it'd be better if the bot could actively introspect and reflect on it than it being haphazardly & accidentally stumbled into after a prompt. That's for a whole other post though.
Azaran on 28/4/2025 at 16:22
Quote Posted by demagogue
I'm thinking at least half of the content you see on social media these days is AI generated.
A ton of my suggested posts on FB are AI trying to pass as real, from pages that seem to be run by bots. It's basically engagement farming, even if the stuff is obviously AI, when they post it publicly, they'll get insane amounts of engagement, both positive (gullible users), negative (people calling it out), and encouraging comments/shares from other bot users. That's the whole point, not entirely deceit, more engagement and traffic
Aja on 28/4/2025 at 17:21
Quote Posted by heywood
Am I the only one who thinks the hype over generative AI is overblown?
No. It's an occasionally useful tool that's been massively overhyped in order to get investors to treat it like it's the next smartphone in the pursuit of never-ending growth. Any time Sam Altman talks about the dangers of AI or how they're on a verge of a breakthrough, you can be certain that it's because they need more money (they lose at (
https://m.economictimes.com/news/international/us/chatgpt-owner-openai-is-losing-money-like-anything-and-this-is-its-biggest-problem/articleshow/117063725.cms) $5 billion every year). If these tools were so life-changingly useful, we'd actually be using them in life-changing ways instead of posting shitty made-up Seinfeld episodes or bad Studio Ghibli impressions. If it was genuinely useful, it wouldn't need the endless hype and constant attempts to convince us it will somehow help us. In and of itself generative AI is not a scam, but the way it's been sold to us absolutely is, to say nothing of the fact that they stole all of its training data (after decades of making the rest of us feel like criminals for using Napster or Pirate Bay) and are basically boiling the oceans to make this all happen. The bubble can't burst fast enough.
Azaran on 28/4/2025 at 17:44
Its capacity for parsing data, reconstruction, and creating new content based on existing data is out of this world.
E.g. you can feed AI a TTLG thread, ask it to create a theater play based on it, and within a minute it will come up with something brilliant, that would have taken a human days or weeks to create.
Would you like to know what a Shakespearean essay about social media would look like? Feed AI the complete works of the author, and watch the magic happen.
You could try and do it yourself, but it might take a while.
Not saying this is good for human creative works, but it's absolutely mind blowing, and will be useful in the field of linguistics, archeology, ancient literature, etc.
There haven't been as many scientific breakthroughs by AI as I expected, but there's been some (
https://www.docwirenews.com/post/the-top-6-ai-breakthroughs-in-healthcare) advances
Aja on 28/4/2025 at 21:47
Quote Posted by Azaran
Would you like to know what a Shakespearean essay about social media would look like?
No, but if I did, would it be worth the amount of energy and money it took to produce it? Personally I'd rather see major tech companies meet their emission targets, and I'd like to be able to buy graphics cards at affordable prices rather than always be at the whims of whoever thinks they can make the most money from a technology that has yet to truly prove its worth and is burning billions of dollars and the planet while they try to figure it out.
Using AI for parsing medical data does seem like an actually sound idea (and maybe the only one that's actually worth the tremendous costs), but that's not the reason companies like Softbank are (
https://www.reuters.com/technology/artificial-intelligence/openai-raise-40-billion-softbank-led-new-funding-2025-03-31/) trying to raise $40 billion to invest in it. They're doing it because AI hucksters keep telling us the revolution is just around the corner despite the fact that AI still completely sucks at almost everything it does apart from low-level coding or summarizing emails, a task which Microsoft seems to think is reason enough to pay monthly subscription fees (the day that I have to use Copilot to summarize emails that other people used Copilot to write is the day I quit my job and go live in the woods like Thoreau). And even then it's not so reliable that you don't have to double-check everything it says to make sure it's not bullshitting.
As for going to see an AI-written play, I wouldn't because I don't care what AI has to say about anything. It's a novelty. Art is about human connection, and without that, it's just window dressing.
PigLick on 29/4/2025 at 02:02
It's not even really about what AI can do, it's how it was created and who created it. There is an inherent bias built into the code of the big ai(GPT, Gemini etc) against race and minority groups, and even the creators themselves admit they don't know how to change that.
OpenAI was originally a not-for-profit and had full transparency on what data it was being trained on.(originally it was 7000 or something free self published novels, that skewed heavily toward vampires and romance lol, then ALL of reddit posts that had 3 or more likes)
Now Microsoft owns like 49 percent of it, they are opaque as fuck and Altman is going to go to public for profit, although there is a fair bit of outcry about that. Pretty much everyone who is working and developing with these mega-corps subscribe to the "effective altruism" concept, which is a total load of bullshit. (Bankman-Fried, anyone?)
I deliberately stayed away from using chatgpt for over a year because I hate billionaires, but out curiosity I tried it recently and even though it is still really just using the transformer tech, it has vastly improved in pretty much every aspect. For example, I used to ask it about jazz theory and it would generally be rubbish, but now if I ask "what scale would you use over a C9#11 chord" it is dead on accurate, even giving me suggestions for other scales, all of which are correct as well.
edit - C lydian dominant if anyone wants to know. (4th mode of G melodic minor)
heywood on 29/4/2025 at 14:41
Quote Posted by Azaran
Its capacity for parsing data, reconstruction, and creating new content based on existing data is out of this world.
E.g. you can feed AI a TTLG thread, ask it to create a theater play based on it, and within a minute it will come up with something brilliant, that would have taken a human days or weeks to create.
Would you like to know what a Shakespearean essay about social media would look like? Feed AI the complete works of the author, and watch the magic happen.
You could try and do it yourself, but it might take a while.
Not saying this is good for human creative works, but it's absolutely mind blowing, and will be useful in the field of linguistics, archeology, ancient literature, etc.
There haven't been as many scientific breakthroughs by AI as I expected, but there's been some (
https://www.docwirenews.com/post/the-top-6-ai-breakthroughs-in-healthcare) advances
I thought about your examples for a moment. If somebody wrote a Shakespearean essay about social media, would I read it? No. What if a forum member created a theater play based on a TTLG thread? Would I watch it? No. If some other forum members whose tastes I'm familiar with highly recommended it, I might give it a try. And I think that illustrates my point. We already live in a content-rich world with more at our fingertips than we can consume in a lifetime. And we've already had (non-generative) AI try to shovel so much of it down our throats that I've turned off autoplay on everything. An explosion of more cheap content is only going to make us more picky about quality and originality and more dependent on each other to discover it amongst the swamp of AI generated noise.
