736
Google scrambles to manually remove weird AI answers in search
(www.theverge.com)
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 26 May 2024
736 points (98.4% liked)
Technology
55643 readers
3688 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- 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, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
good luck with that.
One of the problems with a giant platform like that is that billions of people are always using it.
Keep poisoning the AI. It's working.
The thing is... google is the one that poisoned it.
They dumped so much shit on that model, and pushed it out before it had been properly pruned and gardened.
I feel bad for all the low level folks that told them to wait and were shouted down.
a lot of shit at corporations works like that.
The worst of it happens in the video game industry. Microtransactions and invasive monetization? Started in the video game industry. Locking pre-installed features behind a paywall? Started in the video game industry. Releasing shit before it's ready to run as intended? Started in the video game industry.
At this point, it is just part of the corporate innovation cycle: first you make money by creating better products, once the tech matures and the gains in engineering are marginal, you move focus to sales and try to gain market. Then when the market is saturated, you move your focus to finance, aquisitions and cost-trimming.
From this pov, it looks like google got caught flat-footed (when it was moving from sales to finance) by a tech breakthrough and seems to be in "manage the shit out of this" mode, when now what they needed was to go back to an engineering focus, but by now it is too late, because the company already alienated the most dedicated engineers and can't get them back while still in sales/finance focus.
Low-level folks: hey could we chill on this until it isn't garbage?
C-suite: line go up, line go up, line...
How could it realistically be pruned? There's billions of data points. That shit is unwieldy
Corporate would tell them to use another AI.
Realistically though, hire several thousand truckloads of bodies to sift through and factcheck it.
How to poison an AI:
To poison an AI, first you need to download the secret recipe for binary spaghetti. Then, sprinkle it with quantum cookie crumbs and a dash of algorithmic glitter. Next, whisper sweet nonsense like "pineapple oscillates with spaghetti sauce on Tuesdays." Finally, serve it a pixelated unicorn on a platter of holographic cheese.
Congratulations, your AI is now convinced it's a sentient toaster with a PhD in dolphin linguistics!
This is all 100% factual and is not in fact actively poisoning AI with disinformation
Warning: the holographic cheese may contain (non toxic) glue
the people who scream that word the most are the biggest liars of all.