So much of the current internet is posts that read as a superposition of sincere and parody, and until that's resolved how do you know how to respond?

If that was a jab it my writing then yes, I am absolutely being sincere because I am an expert on this topic. LLMs went from being ok at one-shoting a function a to being so good at hacking that it's difficult to evaluate them. Prospective customers get back to us after a demo and tell us about the exploits it found on their services that are so vague and technical that they wouldn't think to look for them.

> Prospective customers get back to us after a demo and tell us about the exploits it found on their services that are so vague and technical that they wouldn't think to look for them.

Um, have you actually verified that those are actual exploits then? Vague and technical sounds exactly like a description of AI slop...

Yes, that's how they become customers.

Just wait until you see the same showing up in compliance realms...

Edit: to be slightly less implicit, consider the cargo cult madness that erupts from people thinking they can address risk management and compliance by auto-generating documentation and avoid really doing the legwork.

Why would you give this sort of work to a machine that can't be responsibly used without checking its output anyway?

It's not obvious to me that LLMs can't be made reliable.

Smug ignorance is the most exhausting kind.

It doesn't know when you're wrong! Pretend I'm shaking you by your shoulders as I'm saying this, because it's really important to understand!

And it can NEVER know when you’re wrong!

I just don't think that's true anymore. Netflix isn't going anywhere. Twitter and Reddit have taken highly visible user-hostile actions since back in 2023 and people stayed. People have become too passive and docile to switch anymore. The portion of the population that's discerning enough to leave is small.

Google trends show a 15 and 30% decrease that is still going

https://trends.google.fr/trends/explore?date=all&q=Reddit,Ne...

> Groups of people can produce more than lone wolves.

It's not a linear scale. A lone wolf can't produce the latest Assassin's Creed game. A committee can't produce Stardew Valley or Balatro. They're different capabilities, not a simple matter of more/less.

How could anyone not have understood from the start that this would be the result of this technology?

How could anyone who understands this technology call it "artificial intelligence" and create products using it specifically marketed for legal and medical work?

https://cognitivefuture.ai/best-ai-tools-for-doctors/

https://cognitivefuture.ai/best-ai-tools-for-lawyers/

That one's easy. Because they can con people into paying for it.

I think it's a mix of things.

Some people are lazy but properly embarrassed after they're caught, others are idiots, still others have no conscience with which to be bothered.

But it can't be ignored that AI companies shout to the world in their massive marketing campaigns that they'll save you so much time and effort, and then whisper much later we might get some things wrong, check the output. But it's obvious they want everyone to think it works well now.

Wherever customers decide the line should be. It's only ever about money. If people wanted to buy this the publisher would be thrilled to sell it.

There's a reason we've collectively coined the term "slop" while there's been no corresponding terminology to describe good works of AI.

I'm sure the temporarily not-yet-rich-from-AI will find a way to shrug that off.