News for hackers. Stuff that matters.
Re:Google edits Super Bowl ad for AI that featured false inf... by firecall on Sunday February 09, 2025 @07:59AM (#42990396)
“the Google executive Jerry Dischler said this was not a “hallucination” – where AI systems invent untrue information – but rather a reflection of the fact the untrue information is contained in the websites that Gemini scrapes.”

So what you are saying is that Gemini is basically useless?

Good to know. Thanks for clarifying that Google!

Re:Google edits Super Bowl ad for AI that featured false inf... by switch007 on Sunday February 09, 2025 @08:07AM (#42990432)
Don't worry, just ask it to cite its sources so you can verify
Re:Google edits Super Bowl ad for AI that featured false inf... by BugsJustFindMe on Sunday February 09, 2025 @08:05AM (#42990421)
AI search as yet lacks necessary incredulity.
Re:Google edits Super Bowl ad for AI that featured false inf... by manuelmoreale on Sunday February 09, 2025 @08:06AM (#42990429)
“I did not invented some bullshit, I just repeated the bullshit someone else has invented” doesn’t really sound like a great argument.
Re:Google edits Super Bowl ad for AI that featured false inf... by ethbr1 on Sunday February 09, 2025 @08:20AM (#42990502)
87% of the Internet is bullshit.
Re:Google edits Super Bowl ad for AI that featured false inf... by guy234 on Sunday February 09, 2025 @08:37AM (#42990594)
how could you possibly know that?
Re:Google edits Super Bowl ad for AI that featured false inf... by mauvehaus on Sunday February 09, 2025 @08:52AM (#42990687)
Re:Google edits Super Bowl ad for AI that featured false inf... by fn-mote on Sunday February 09, 2025 @08:40AM (#42990608)
That was the point…
Re:Google edits Super Bowl ad for AI that featured false inf... by jrootabega on Sunday February 09, 2025 @08:19AM (#42990492)
No, they're saying that Google searching in general is often useless.
Re:Google edits Super Bowl ad for AI that featured false inf... by UncleMeat on Sunday February 09, 2025 @09:13AM (#42990803)
If I do a google search I can inspect the source of a response for trustworthiness and I can inspect several other sources that also appear in the results for agreement.

With an AI query I get one response back and although people are trying to teach their AI tools to cite sources, this is not universal.

Re:Google edits Super Bowl ad for AI that featured false inf... by g-b-r on Sunday February 09, 2025 @09:02AM (#42990746)
Checking your web search sources and credibility is a lot easier
Re:Google edits Super Bowl ad for AI that featured false inf... by crazygringo on Sunday February 09, 2025 @08:38AM (#42990597)
That's quite a leap there from "sometimes wrong" to "basically useless"...
Re:Google edits Super Bowl ad for AI that featured false inf... by rdtsc on Sunday February 09, 2025 @10:40AM (#42991307)
> quite a leap there

A healthy hop I would say, keeping in mind that this was showing off the best side of it in a commercial. So it just did a Google search, got a wrong factual result, the LLM couldn’t verify it was bogus, regurgitated it as is, and the executive piped in with “acshually, it’s just a plain old search result” somehow not realizing that just makes them look even worse.

Re:Google edits Super Bowl ad for AI that featured false inf... by WhyNotHugo on Sunday February 09, 2025 @09:11AM (#42990799)
If a model often produces information that is blatantly wrong, then you need to check ALL of its outputs. If you're going to have to double-check all information that it provides, you might as well skip using it entirely and search for the information directly.
Re:Google edits Super Bowl ad for AI that featured false inf... by crazygringo on Sunday February 09, 2025 @09:59AM (#42991058)
You're missing the part where searching for the information directly might take hours or even days.

You're looking for black-and-white truth, while the real world is actually more interested in efficiency.

I can spend 2 days scouring documentation and forum posts and experimenting to get ffmpeg or Matplotlib to produce exactly the results I want. Or I can just ask ChatGPT and check if its code works, and if it doesn't maybe spend 10 minutes correcting it so it does, or refining the prompt so it does.

And so you're also missing the part where verifying the correctness of output is very often many orders of magnitude faster than coming up with the output in the first place.

Re:Google edits Super Bowl ad for AI that featured false inf... by nolist_policy on Sunday February 09, 2025 @09:49AM (#42991003)
LLMs produce wrong results. News at 11.
Re:Google edits Super Bowl ad for AI that featured false inf... by loveparade on Sunday February 09, 2025 @08:43AM (#42990630)
To be fair, do you believe a human would do better? If you hired someone to create content for your cheese website they would probably do the same thing. They'd search google and copy statistics without triple-checking all of them.

I don't think it's fair to frame this is an AI issue. The internet is simply full of misinformation. Should AI outperform the average human and detect misinformation? Maybe. But I don't think that's part of the current value prop, at least not for "non-reasoning" models. If you use an LLM you should be aware that what you're getting is whatever is most commonly found on the web. If that's not what you want, don't use an LLM.

Re:Google edits Super Bowl ad for AI that featured false inf... by stefan_ on Sunday February 09, 2025 @08:46AM (#42990644)
Amazing, they got the "actually" guy for a PR person. That's still shit Jerry!
Re:Google edits Super Bowl ad for AI that featured false inf... by Rebelgecko on Sunday February 09, 2025 @04:11PM (#42994024)
For better or worse Jerry isn't a PR person
Re:Google edits Super Bowl ad for AI that featured false inf... by frenchtoast8 on Sunday February 09, 2025 @08:48AM (#42990660)
Not just the "50 to 60 percent" statistic, it looks like the entire paragraph in the ad is word for word identical to a preexisting cheese shop site. So either Gemini's "generation" is just searching the web and doing verbatim copy paste, or Google copied it themselves which would be a new low for Google.
Re:Google edits Super Bowl ad for AI that featured false inf... by coliveira on Sunday February 09, 2025 @09:07AM (#42990778)
A lot of the answers given by Gemini are just a verbatim copy of parts of web sites. I have seen this time after time when comparing the AI results and text that I find in the referenced web sites.
Re:Google edits Super Bowl ad for AI that featured false inf... by palmotea on Thursday February 13, 2025 @04:54PM (#43041877)
> it looks like the entire paragraph in the ad is word for word identical to a preexisting cheese shop site

The cheese shop should sue Google for copyright infringement. That's also blatant plagiarism, and Google should get an F for the course.

Re:Google edits Super Bowl ad for AI that featured false inf... by ssharp on Sunday February 09, 2025 @08:59AM (#42990730)
> “the Google executive Jerry Dischler said this was not a “hallucination” – where AI systems invent untrue information – but rather a reflection of the fact the untrue information is contained in the websites that Gemini scrapes.”

This is a huge fail for Gemini. Google, of all companies, should know the incentives to distort information aggregators for monetary gain (just look at how bad search results are these days). It's 100% expected that people will try to game LLMs the same way. It's entirely dependent on the LLMs to counteract this.

Re:Google edits Super Bowl ad for AI that featured false inf... by drawkward on Sunday February 09, 2025 @09:31AM (#42990909)
Maybe the goal is not accuracy, but market capture?
Re:Google edits Super Bowl ad for AI that featured false inf... by coliveira on Sunday February 09, 2025 @09:03AM (#42990750)
Bingo. They admitted what everyone suspected: they are just scrapping BS from BS websites and spilling it back as something created by an "intelligent agent".
Re:Google edits Super Bowl ad for AI that featured false inf... by qwertox on Sunday February 09, 2025 @12:21PM (#42991984)
> but rather a reflection of the fact the untrue information is contained in the websites that Gemini scrapes

It was a verbatim copy of the text, which raises the question about copyright. I guess "Wisconsin Cheese Mart" could probably sue Google, but more importantly, this would also mean that it is capable of making verbatim copies of the books contained in Anna's Archive.

That is, assuming that Google also did torrent the books like Meta did, which is very likely.

Funny how they call it "reflection", and not "copy". Maybe that's the way to win in courts: I didn't copy that movie, that was just a reflection of it.

Wouldn't this user of Gemini, who would copy Gemini's response to use this text in their ads, potentially get into copyright troubles with "Wisconsin Cheese Mart"?