Hacker News Reader

Comment by jqpabc123 | Hacker News Reader

AI doesn't know or care if the data is fake.

I wonder if Googles products will eventually be the least vulnerable to this after spending so many years optimizing how they filter quality content from bad content.

Or maybe they could have listed well on Google, too, if they had splurged on ads and SEO...

The page demonstrates that Google specifically delisted them.

That said, I expect that OpenAI will eventually have its own search infrastructure that rivals Google's, and the overall system will get smarter. Whatever Google did to recognize this site as trash, OpenAI will do as well.

Considering Google’s product was the one who told people to eat rocks, put glue on pizza, and their search results have been dominated by SEO spam for years, I’m not optimistic.

This is dinosaur era LLMs. Might as well complain about use of shockwave flash in production while you’re at it.

Reminds me of the people whining about Wikipedia. AI can know or care if the data is fake, it just needs to be programmed to cross check or rate the source. You can even tell it to do that before accepting the data.

These are not parallel situations. Wikipedia is far more accurate and consistent, and a lot of the whiners whine because their personal politics make it hard to square the information on the page.

Often real people don't either, plenty of fraud or selective reporting out there... but with AI we can also launder it so that the people who do care about fake data don't know any better.

Yep, and the results are presented with the voice of a confident expert but it is actually missing the epistemological layer

You can prompt it to fact check. Getting tired of these slippery slope arguments that never pan out in the hopeless way people think (printing press/cotton gin/steam engine/computers/calculators/etc. will take all the jobs!!)... but I don't think it will ever stop, it seems inherent to human behavior or something.

"History shows that when a slope appears, regulation steps in, technology evolves to solve the problem, or the culture shifts to reinterpret the tool. In almost every case, the feared bottom of the slope was never reached because humans constantly built ramps or bridges along the way."

in which case it sees it's been cited, that it's the "source of truth" and approves the fake data.

It can certainly make judgements on the data and how it was obtained. You just need to specify the kind of data you will accept (or this can be hard coded)

Fact check against what? Other facts it can find online that manipulate its behaviour into citing them?

How do humans fact check?

Mostly by way of confirmation bias.

if it works often enough to be a good heuristic, it's not really a bias until somebody uses it adversly

It is used adversely all the time by all sorts of propaganda. And no, you're not a special enlightened snowflake whose priors are so fine-tuned that your confirmation bias only lets true facts through.

> Be kind. Don't be snarky. Converse curiously; don't cross-examine. Edit out swipes.

https://news.ycombinator.com/newsguidelines.html

You cannot reliably do it, no.

I've had AI lie about fact checking and can't really trust it anymore.

The fact that AI can be duped more easily than a human while sounding more authoritative and being entrusted with more and more critical tasks is really not something you should feel compelled to defend or relativize. It's a problem and the industry doesn't seem to have a good fix.

> the feared bottom of the slope was never reached

Hard disagree. Sometimes it's avoided, any other time we just go into cope mode. We discover on the ride down that the bottom isn't so bad. Turns out our collective control mechanisms didn't fail ourselves if the initial assessment was wrong.

Reply on news.ycombinator.com