Google scrambles to manually remove weird AI answers in search • Hacker News Board

Re: Google scrambles to manually remove weird AI answers in search

#11

Post by caesil » Sat, May 25, 2024, 4:06 PM UTC

Who knows how many of these are fake. People have been dropping inspect-element-manipulated screenshots all over twitter.

https://www.nytimes.com/2024/05/24/technology/google-ai-over...

> A correction was made on May 24, 2024: An earlier version of this article referred incorrectly to a Google result from the company’s new artificial-intelligence tool AI Overview. A social media commenter claimed that a result for a search on depression suggested jumping off the Golden Gate Bridge as a remedy. That result was faked, a Google spokeswoman said, and never appeared in real results.

that screenshot was tweeted by @allgarbled. ten minutes before, they tweeted:

>free engagement hack right now is to just inspect element on the google search AI thing and edit it to something dumb. hurry up, this deal won’t last forever

Re: Google scrambles to manually remove weird AI answers in search

#12

Post by ethbr1 » Sat, May 25, 2024, 4:08 PM UTC

caesil wrote: post #11

Who knows how many of these are fake. People have been dropping inspect-element-manipulated screenshots all over twitter. https://www.nytimes.com/2024/05/24/technology/google-ai-over... > A correction was made on May 24, 2024: An earlier version of this article referred incorrectly to a Google result from the company’s new artificial-intelligence tool AI Overview. A social media commenter claimed that a result for a…

I'd say the broader issue here is a lack of transparency into results.

If Google is sending bad results, who can prove that?

Re: Google scrambles to manually remove weird AI answers in search

#13

Post by YetAnotherNick » Sat, May 25, 2024, 4:08 PM UTC

CatWChainsaw wrote: post #10

How hard can it possibly be to just turn off the entire AI-generated overview functionality given that it just got introduced...

It seems to be turned off for me. And I was in beta testing for a month. Or maybe they are figuring out who is doing weird searches and turning off for them.

In any case this thing is just hilarious. Just right after their AI painted historical figures as black.

Re: Google scrambles to manually remove weird AI answers in search

#14

Post by rchaud » Sat, May 25, 2024, 4:09 PM UTC

dazc wrote: post #2

Manually removing rogue AI results is kind of ironic isn't it?

Pay no attention to the Accenture contractors behind the curtain!

Re: Google scrambles to manually remove weird AI answers in search

#15

Post by pilooch » Sat, May 25, 2024, 4:09 PM UTC

So Google hasn't used an LLM to generate and test weird queries ? This is not putting the bar very high for the whole industry... There'd be so much to gain from a clean deployment... Either it hard, either it is a rush. As a machine learnist, I believe it's actually impossible, by design of the autoregressive LLM. This race may we'll be partially to the bottom.

Re: Google scrambles to manually remove weird AI answers in search

#16

Post by szundi » Sat, May 25, 2024, 4:10 PM UTC

moomoo11 wrote: post #4

Just focus on making useful software to improve people’s lives. Holy fuck the last five years feel like such a waste.

Management realized that they are not good enough to sustain progress, so they humbly allocate resources to the next generation: AI

Re: Google scrambles to manually remove weird AI answers in search

#17

Post by JCM9 » Sat, May 25, 2024, 4:10 PM UTC

The fall of Google’s reputation on ML is nothing short of spectacular. They went from having a near untouchable reputation as being far ahead of any other large tech company on ML to total shambles in a year. Everything they’ve released has been a complete popcorn worthy dumpster fire from faked demos, to racist models that try and pretend white people don’t exist, to this latest nonsense telling me put glue on my pizza.

What the heck happened? Or was their reputation always just more hype than substance?

Re: Google scrambles to manually remove weird AI answers in search

#18

Post by CaptainOfCoit » Sat, May 25, 2024, 4:10 PM UTC

pilooch wrote: post #15

So Google hasn't used an LLM to generate and test weird queries ? This is not putting the bar very high for the whole industry... There'd be so much to gain from a clean deployment... Either it hard, either it is a rush. As a machine learnist, I believe it's actually impossible, by design of the autoregressive LLM. This race may we'll be partially to the bottom.

> So Google hasn't used an LLM to generate and test weird queries ?

What about simple manual testing? Seems to have skipped QA completely, automated or not.

Re: Google scrambles to manually remove weird AI answers in search

#19

Post by rchaud » Sat, May 25, 2024, 4:11 PM UTC

CatWChainsaw wrote: post #10

How hard can it possibly be to just turn off the entire AI-generated overview functionality given that it just got introduced...

very hard indeed, if you're optimizing for favourable opinions from Wall St analysts come earnings time .

Re: Google scrambles to manually remove weird AI answers in search

#20

Post by pilooch » Sat, May 25, 2024, 4:14 PM UTC

CaptainOfCoit wrote: post #18

pilooch wrote: post #15

So Google hasn't used an LLM to generate and test weird queries ? This is not putting the bar very high for the whole industry... There'd be so much to gain from a clean deployment... Either it hard, either it is a rush. As a machine learnist, I believe it's actually impossible, by design of the autoregressive LLM. This race may we'll be partially to the bottom.

> So Google hasn't used an LLM to generate and test weird queries ? What about simple manual testing? Seems to have skipped QA completely, automated or not.

The adversarial surface to the LLM remains enormous, manual cannot handle it.