this post was submitted on 25 May 2026
454 points (96.5% liked)
Technology
84920 readers
3619 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I don't understand how these companies want to seem and think they are so smart by choosing new niche data (scraped) to train AI in a bid to try and make it "smart"....
Has any other living being become "smart" by only ingesting information directly from the Internet? You can train other animals to perform many tasks and can probably say they are smart when they perform them as expected. I doubt any of the training methods is to tape headphones, a screen and sometimes a microphone to their faces forever (I kinda don't wanna know if this false 😶).
The best example we have, is ourselves, and even though we use the Internet, babies are not taught how to walk and talk by only interacting with the Internet.
I feel like I might be saying too much, but I think the best AI we're gonna get is to unplug it from the Internet, and then fucking raise it for 20 years like a normal, super fast-thinking child prodigy. Then just make copies of that and train further by having it go to school for the things needed.
That's a very naive simplification of the AI training process. You start with that, then pay people pennies in a developing nation to produce hand crafted training data, resulting it using stupid words like delve and whimsical entirely too much.
Merely training on internet content with no RLFH training results in probable gibberish like that of GPT-2
LMFAO