pennomi
@pennomi@lemmy.world
- Comment on ChatGPT "Absolutely Wrecked" at Chess by Atari 2600 Console From 1977 2 days ago:
Spatial reasoning has always been a weakness of LLMs. Other symptoms include the inability to count and no concept of object permanence.
- Comment on Scientists discover that feeding AI models 10% 4chan trash actually makes them better behaved 3 days ago:
I mean, it still could be. But LLMs are not that AGI we’re expecting.
- Comment on Scientists discover that feeding AI models 10% 4chan trash actually makes them better behaved 3 days ago:
“Don’t believe that marketing department“ is one of those things everybody needs to learn at some point in their life.
- Comment on 🆓🆒🆗 3 days ago:
Congratulations for being one of the few people who made a POV meme correctly!
- Comment on Fractured Blooms is a cosy game about tending your garden and-- oh god that's a lot of blood 6 days ago:
It was announced at the Summer Game Fest, where presenter Geoff Keighly described the game as “Doki Doki Literature Club meets Silent Hill”. There’s no release date, but here’s a trailer.
That’s a pretty high bar. If they can pull it off it’ll be legendary.
- Comment on Google confirms more ads on your paid YouTube Premium Lite soon 1 week ago:
Is it in decline? I mean, I want to believe it, but I haven’t seen any hard data on that.
- Comment on Pro-AI Subreddit Bans 'Uptick' of Users Who Suffer from AI Delusions 1 week ago:
I don’t think I suggested it wasn’t worrisome, just that it’s expected.
If you think about it, AI is tuned using RLHF, or Reinforcement Learning from Human Feedback. That means the only thing AI is optimizing for is “convincingness”. It doesn’t optimize for intelligence, anything seems like intelligence is literally just a side effect as it forever marches onward towards becoming convincing to humans.
“Hey, I’ve seen this one before!” You might say. Indeed, this is exactly what happened to social media. They optimized for “engagement”, not truth, and now it’s eroding the minds of lots of people everywhere. AI will do the same thing if run by corporations in search of profits.
Left unchecked, it’s entirely possible that AI will become the most addictive, seductive technology in history.
- Comment on Pro-AI Subreddit Bans 'Uptick' of Users Who Suffer from AI Delusions 1 week ago:
Turns out it doesn’t really matter what the medium is, people will abuse it if they don’t have a stable mental foundation. I’m not shocked at all that a person who would believe a flat earth shitpost would also believe AI hallucinations.
- Comment on Uploading The Human Mind Could Become a Reality, Expert Says 1 week ago:
Sensationalist headline about science fiction.
No shit it’s “possible” despite being hopelessly beyond today’s technology.
- Comment on It’s Time To Go Back to Web 1.0 2 weeks ago:
Yeah that stuff bugs the shit out of me. It’s like people just discovered they can send HTML to a browser.
Skipping that whole loop would be far more efficient but it’s not cool to do that any more.
- Comment on If you didn't know the context, 'villager breeder tutorial 1 18' would be a very sketchy thing to search on school computers 2 weeks ago:
Rimworld: What is the most profitable product to make from human leather?
- Comment on Subnautica Is Coming To Android/iOS 2 weeks ago:
- Exclusive mobile UI with touch controls
What a weird way to phrase that.
- Comment on Life hack 2 weeks ago:
Sometimes it’s about the class of plane and the weather on the specific route you’re taking.
- Comment on DOOM: The Dark Ages Has Reportedly Sold Less Than 1 Million Copies 2 weeks ago:
Depends on your location in the world. Which shows just how arbitrary the price actually is.
- Comment on Valve CEO Gabe Newell’s Neuralink competitor is expecting its first brain chip this year 2 weeks ago:
I suspect you are all too right.
- Comment on Valve CEO Gabe Newell’s Neuralink competitor is expecting its first brain chip this year 2 weeks ago:
I am self-employed. So myself, I guess.
- Comment on Valve CEO Gabe Newell’s Neuralink competitor is expecting its first brain chip this year 2 weeks ago:
It’s risk/reward. If brain chips made me twice as productive or intelligent, I’d probably tolerate a lot more risk than if it was just a way to check my Instagram notifications without pulling out my phone.
- Comment on Valve CEO Gabe Newell’s Neuralink competitor is expecting its first brain chip this year 2 weeks ago:
Is a killer a good person?
Well yeah, sometimes! It is very context dependent.
- Comment on The Copilot Delusion 2 weeks ago:
Not all AI is made equally. There are Apache licensed LLMs.
- Comment on Black Mirror AI 2 weeks ago:
It absolutely doesn’t. The only model that has “gone nuts” is Grok, and that’s because of malicious code pushed specifically for the purpose of spreading propaganda.
- Comment on It's Breathtaking How Fast AI Is Screwing Up the Education System 3 weeks ago:
AI, please write my thesis in the style of Shakespeare. Good luck detecting THAT as AI writing.
- Comment on It's Breathtaking How Fast AI Is Screwing Up the Education System 3 weeks ago:
2 person classes would be a dream compared to the overburdened 30+ person classes of today. You get half of a private tutor? Hell yeah.
- Comment on Time Stop 3 weeks ago:
See that’s the clever thing about this joke. They’re playing off of the regular victim blaming trope “it’s your fault for being dressed like that” but he actually says “for being careless walking”.
The implication was exactly what you thought, until it’s expertly subverted because it turns out he’s literally saving her from danger.
- Comment on Israel will 'take control of all' of Gaza, PM says 3 weeks ago:
- Comment on Like it ever gonna happen 4 weeks ago:
Yeah no worries, I was just being silly by intentionally reading your comment in an overly literal way. I don’t think anyone actually believed you were implying it was a women-only issue.
- Comment on The joy of a family that values education celebrating the graduation of their son 4 weeks ago:
Congration!
- Comment on Like it ever gonna happen 4 weeks ago:
Hmm yes, that’s definitely a women only problem that doesn’t also affect me because I’m a man.
- Comment on Researchers unveil LegoGPT, an AI model that designs physically stable Lego structures from text prompts and currently supports eight standard brick types 4 weeks ago:
Nobody can tell you, and if they say they do know, they’re selling snake oil.
But when it happens, shit is going to change fast, for better or for worse.
- Comment on Cloudflare CEO warns AI and zero-click internet are killing the web's business model 4 weeks ago:
Federation, mostly. Places like Lemmy and Mastodon aren’t obsessed with optimizing ad spend, so there’s a lot lees incentive to enshittify.
- Comment on Not difficult to understand 5 weeks ago:
Haha, as if anyone actually asks that. A gap gets you automatically rejected by the employment AI, long before humans see it.
