Technology

Do you think Google will recommend microwaving your iPhone to recharge it's battery at some point?

I want AI answers that end saying that in 1998, The Undertaker threw Mankind off Hell In A Cell, and plummeted 16 ft through an announcer's table.

oh gods what happens when the ai discovers the poop knife

How the fuck did none of those expensive ties at Google see this happening? Have your AI devour the dumbest shit on the internet, then unleash it to human centipede that diarrhea into the mouths of their users. "Elite" is a fucking joke, ya'll are just as fucken stupid as the rest of us.

So, basically shitposting poisons AI training. Good to know 👍

Maybe try the recipe before you talk shit, you scaredy cats.

I love that my almost 2 decades of shitposting will be put to... use?

Lot of people not liking 404 Media, but this is the kind of reporting I want. Point out what's going wrong. Bring it to a conversation without a lot of skew. Fucking show the general reading audience how they are being fleeced by whomever. Didn't Vice do this at one point?

Reddit, and by extension, Lemmy, offers the ideal format for LLM datasets: human generated conversational comments, which, unlike traditional forums, are organized in a branched nested format and scored with votes in the same way that LLM reward models are built.

There is really no way of knowing, much less prevent public facing data from being scraped and used to build LLMs, but, let's do an thought experiment: what if, hypothetically speaking, there is some particularly individual who wanted to poison that dataset with shitposts in a way that is hard to detect or remove with any easily automate method, by camouflaging their own online presence within common human generated text data created during this time period, let's say, the internet marketing campaign of a major Hollywood blockbuster.

Since scrapers do not understand context, by creating shitposts in similar format to, let's say, the social media account of an A-list celebrity starring in this hypothetical film being promoted(ideally, it would be someone who no longer has a major social media presence to avoid shitpost data dilution), whenever an LLM aligned on a reward model built on said dataset is prompted for an impression of this celebrity, it's likely that shitposts in the same format would be generated instead, with no one being the wiser.

That would be pretty funny.

Again, this is entirely hypothetical, of course.

There’s an old adage in computing which really applies here:

Garbage in, garbage out.

This is why you don't train a bot on the entire Internet and then use it to offer advice. Even if only 1% of all posts are dangerously ignorant . . . that's a lot of dangerous ignorance.

Fortunately, this particular piece of bad advice is unlikely to poison any fool who goes through with it, since PVA glue is not considered an ingestion hazard, but "non-toxic" doesn't mean "edible", it just means "not going to poison you when used in the intended manner". "Non-toxic" can still be quite dangerous if you mistake something intended as linoleum pigment for a dessert topping.

I've been trying out SearX and I'm really starting to like it. It reminds me of early Internet search results before Google started added crap to theirs. There's currently 82 Instances to choose from, here

https://searx.space/

Thr problem the AI tools are going to have is that they will have tons of things like this that they won't catch and be able to fix. Some will come from sources like Reddit that have limited restrictions for accuracy or safety, and others will come from people specifically trying to poison it with wrong information (like when folks using chat gpt were teaching it that 2+2=5). Fixing only the ones that get media attention is a losing battle. At some point someone will get hurt or hurt others because of the info provided by an AI tool.

You: "How do I make a pizza?"

Reddit-Bot: "Did you know the first recorded Bitcoin transaction was 10,000 bitcoins for two pizzas? Pizza is much cheaper now so just go buy it."

I've used an LLM that provides references for most things it says, and it really ruined a lot of the magic when I saw the answer was basically copied verbatim from those sources with a little rewording to mash it together. I can't imagine trusting an LLM that doesn't do this now.

Here's Google suggesting suicide!

That's a great read if you are only trying to film a commercial or promotion and no one is going to eat it. But then it doesnt matter if its non toxic i suppose.

At least i remember a video a long time ago, perhaps on an episode of how its made, that white glue is used to help get the stretchy cheese pull