Live data from Hacker News

Generative AI could make search harder to trust

wired.com

201–210 of 255 posts

Re: Generative AI could make search harder to trust

#201

Earlier quoted context omitted.

People are vastly everestimating how unique this problem of hallucinations is. It seems to me it relies mostly on discounting just how much we've already had to deal with this same problem in humans over the millenia. The problem of proliferation of bad information might be getting worse, but this isn't native to generative AI. The entire informational ecosystem has to deal with this. GPTs compound the issue, but as…

The thing is when you call a human on bullshit, they usually can't back it up well enough to pass the smell test. When you call an AI on bullshit it can instantly fabricate plausible, credible seeming sources/evidence. A human's lie is different than an AI's hallucination, since it's still based on (distorting) the truth, whereas the hallucination is based on an invented reality (yes I know it's applied statistics an…

Evidence can easily be cross referenced and if the link LLM gives is dead it means it's lying.

Re: Generative AI could make search harder to trust

#203

I actually experienced this the other day. Bought the new Baldur's Gate and was wondering what items to keep or sell (don't judge me, I'm a pack rat in games!) I had found some silver ingots. The top search result for "bg3 silver ingot" is a content farm article that very confidently claims you can use them at a workbench in Act 3 to upgrade your weapons. Except this is a complete fabrication: silver ingots exist onl…

Probably in the future people will only trust sources of info that can't be monatised. If you want to know the answer to a game question you just got to the reddit or discord and ask, since there is no point autogenerating crap for discord when you can't put ads next to it and the mods can remove you.

I guess we need to again (and again) establish that having good information isn't necessarily cheap: could be a high quality vendor, policing a forum, maintaining search engine integrity, etc.

Re: Generative AI could make search harder to trust

#205

You should never have had been trusting what you find on the web, let alone other media. I hope widespread usage of generative AIs including deepfakes will finally force the masses to start thinking more critically.

Also it might reinforce cognitive biais as pure truths to a lot of people who rely only on themselves if no other information is reliable

Re: Generative AI could make search harder to trust

#206

It’s especially terrifying that misinformation compounds multiplicatively with AI because it happens in 2 layers - once at the retrieval layer (where AI-generated content is worsening the problem of bad SEO content) and again at the retrieval augmented generation (RAG) LLM layer. (shameless plug) At Metaphor ( https://platform.metaphor.systems/ ), we’re building a search engine that avoids SEO content by relying on h…

+1. Increasing training data quality should also hopefully help with hallucination. But becomes increasingly hard in a world where more and more online content is crap.

I'm not an expert in training LLMs, but I've heard that some people use reinforcement algorithms to train and align LLM behaviors with human preferences. When it comes to designing a loss function for training, I wonder if it's possible to assign an extremely high loss value to hallucinated content during training. This approach might encourage the model to refrain from generating inaccurate content.

Re: Generative AI could make search harder to trust

#207
I think the insufficient accuracy in the output of LLMs is going to lead them to be a lot more niche then the current hype is hoping for. I think most people care that someone is taking accountability for what they are reading - not that it is necessarily correct but at least that someone thinks it is correct (and that someone can be taken to task if it is inaccurate).

If LLM usage in media becomes widespread, I'd pay for a service that identifies and hides the LLM shit for me.

Re: Generative AI could make search harder to trust

#208

Earlier quoted context omitted.

That's fine. I genuinely wouldn't want to continue working in an industry that worked like that anyway, so I'd just quit and keep on programming with my own projects. So that luxury will last as long as I want it to.

I've come to the same conclusion. I became a programmer because in school I loved programming and saw myself loving the job, and I do. If the job changes that drastically I'll just have to quit and find something else.

But what if you have to use an LLM at your new job?

Re: Generative AI could make search harder to trust

#209

Earlier quoted context omitted.

Generative Wayback Machine

Honestly, I would love to see a LLM based plugin that would take a website, remove all the tracking garbage and filler, and just give me back a no-frills static html+CSS site that looks like it was made in 1995. "Oh, you want a guide to writing your own loss function for Tensorflow? Here's an FAQ that could have existed on comp.lang.python3.tensorflow"

More unnecessary technology doesn't solve the problem of too much unnecessary technology.

Re: Generative AI could make search harder to trust

#210

Earlier quoted context omitted.

Yes. I love black text on white background. A rare find these days. Browsing today is like: “You ask for a spaghetti recipe and the page tell you the whole history of civilization.”

Don't forget how Google will now drop search terms if it thinks you mean something else, or add unrelated synonyms to your search (presumably to "help" folks who aren't good at writing queries)

At least it still lets you force whatever you write in...

Allegro (big polish auction/e-commerce site) in their mobile app will unconditionally rewrite the search terms instead of showing you no results

Post reply on HN