this post was submitted on 27 May 2024
1103 points (98.1% liked)

Technology

68130 readers
3547 users here now

This is a most excellent place for technology news and articles.


Our Rules


  1. Follow the lemmy.world rules.
  2. Only tech related news or articles.
  3. Be excellent to each other!
  4. Mod approved content bots can post up to 10 articles per day.
  5. Threads asking for personal tech support may be deleted.
  6. Politics threads may be removed.
  7. No memes allowed as posts, OK to post as comments.
  8. 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.
  9. Check for duplicates before posting, duplicates may be removed
  10. Accounts 7 days and younger will have their posts automatically removed.

Approved Bots


founded 2 years ago
MODERATORS
 

You know how Google's new feature called AI Overviews is prone to spitting out wildly incorrect answers to search queries? In one instance, AI Overviews told a user to use glue on pizza to make sure the cheese won't slide off (pssst...please don't do this.)

Well, according to an interview at The Vergewith Google CEO Sundar Pichai published earlier this week, just before criticism of the outputs really took off, these "hallucinations" are an "inherent feature" of  AI large language models (LLM), which is what drives AI Overviews, and this feature "is still an unsolved problem."

(page 2) 50 comments
sorted by: hot top controversial new old
[–] [email protected] 15 points 10 months ago (3 children)

How about turn it the fuck off since it sucks and eventually will kill someone.

load more comments (3 replies)
[–] [email protected] 8 points 10 months ago (1 children)

These models are mad libs machines. They just decide on the next word based on input and training. As such, there isn’t a solution to stopping hallucinations.

load more comments (1 replies)
[–] [email protected] 27 points 10 months ago (1 children)

The model literally ate The Onion, and now they can't get it to throw it back up.

[–] [email protected] 4 points 10 months ago

I love this wording, because it’s so true

[–] [email protected] 5 points 10 months ago

Looks like Google stopped the AI feature. No more AI suggestions at the top of the page after searching for something.

[–] [email protected] 5 points 10 months ago (1 children)

I'm curious, are these hallucinations very prevalent? I'm outside under US so haven't seen the feature yet. But I have noticed that practically every article references the same glue incident.

So I'm not sure if the hallucinations are happening all the time, or everyone is just jumping on a handful of mistakes the AI made. If the latter, the situation reminds me of how every single accident involving a Tesla was reported on back in the day.

[–] [email protected] 7 points 10 months ago

It will confidently report inaccurate information. It's usually not so hilariously wrong, but it's still wrong.
For example, I was talking with someone about what constituents a "fruit" botanically, and I searched "are beans fruit", and it confidently told me that beans are not a fruit, botanically speaking, because they're a legume. It seems to have adapted, but that's a good example of a "small wrong" that's not uncommon at all.

[–] [email protected] 20 points 10 months ago

Then fucking turn it off

[–] [email protected] 43 points 10 months ago (2 children)

This is so wild to me... as a software engineer, if my software doesn't work 100% of the time as requested in the specification, it fails tests, doesn't get released and I get told to fix all issues before going live.

AI is basically another word for unrealiable software full of bugs.

[–] [email protected] 3 points 10 months ago

And therein lies the difference between engineers and business people. And look which ones are usually in charge.

[–] [email protected] 3 points 10 months ago

Depends on how strict you are about the tests. Google is obviously satisfied if the first live iteration of a product doesn't kill more than 5% of the users.

[–] [email protected] 24 points 10 months ago

I mean yeah... if he had a solution they would be actually have the revolutionary AI tool the tech writers write about.

It's kinda written like a "gotcha" but it's really the fundamental problem with AI. We call it hallucinations now but a few years ago we just called it being wrong or returning bad results.

It's like saying we have teleportation working in that we can vaporize you on the spot but are just struggling to reconstruct you elsewhere. "It's halfway there!"

Until the AI is trustworthy enough to not require fact checking it afterwards it's just a toy.

[–] [email protected] 14 points 10 months ago* (last edited 10 months ago)

Maybe Google should put a disclaimer... warning people it's not 100% accurate. Or.. just take down the technology because clearly their AI is chit tier.

[–] [email protected] 5 points 10 months ago (2 children)

I have a solution! Employ a human to verify the work of AI, perhaps you need more than one with all the junk AI might produce. Maybe you will even need an entire department to do that, and maybe you should just not use AI.

[–] [email protected] 3 points 10 months ago

Whoah, whoah, whoah. Hire people!? We’re here to take all the money. Not spend money. — Google

[–] [email protected] 2 points 10 months ago

Sadly technology companies are going to slowly force us into using AI technology. For example, Apple's iOS update 18 will incorporate AI.

[–] [email protected] 20 points 10 months ago (1 children)

Google really thinks they have no competition. Like, fuck, even Bing improved a lot with time.

[–] [email protected] 2 points 10 months ago

When MS was pushing the Bing challenge with TV ads, it really was quite close. I did the challenge and Google "won" (only) 3/5 of my test searches.

Of course Bing already had a hilariously incorrect "AI" interfering with the first page of results for a week or two before Google decided to further fsck up their search with LLM response generation.

[–] [email protected] 30 points 10 months ago (2 children)

I have a solution: stop using their search engine to begin with and slowly replace everything else google you use.

[–] [email protected] 2 points 10 months ago

Wish more and more people knew about Linux phones and degoogled phones.

[–] [email protected] 12 points 10 months ago (2 children)

Just use a VPN, you will immediately stop using Google after having to identify your umpteenth fire hydrant...

[–] [email protected] 5 points 10 months ago (1 children)

Those boxes that only contain 1/2 of a motorcycle mirror or handle- do we say they contain motorcycle?

[–] [email protected] 1 points 10 months ago (1 children)

Literally doesn't matter; sometimes it's checking for indicators in your mouse movement, sometimes it's checking your cookies... It never actually cares if you've correctly identified or not.

load more comments (1 replies)
[–] [email protected] 2 points 10 months ago

Are you me? I also switched after getting a VPN.

load more comments
view more: ‹ prev next ›