Problem wide open. Microsoft already thought of this solution at least 25 years ago.

They implemented warnings by interrupting the code, opening a pop-up window with two options: proceed or bloc?. I'll give you a guess how that panned out.

There is only one outcome when users are repeatedly interrupted for security reasons. They learn to press yes without even reading the message, while being annoyed by the interruption. Black hats love that.

Child, you do not live in a vacuum. You live in a country with laws. These laws apply to you. Ignoring them will be bad for you. Don't do it.

There are two issues here.

1) Ads are evil, they are a form of propaganda.

2) LLMs are ideally suited to be ad machines, unconstrained from reality.

OpenAI is desperate for revenue, to claw itself out of the gigantic debt hole that Altman has created. Whether this will work is unlikely, but the advertising move will produce much revenue initially. It will also make it obvious that LLMs are a waste of time, eventually, and a form of spam that should be outlawed.

That was a crime problem, not a prohibition problem. Turns out when you let criminals run around with impunity for years, you get bad behaviour. But if you shoot them or prosecute them promptly, even if it's for tax evasion, things get better. Who knew?

No, no, and no. Do not confuse patents and copyrights. Two entirely different kettle of fish. Both are elements of the intellectual property stack. Yet they work entirely differently.

If you want change, maybe educate yourself on IP laws, then work to change these laws. Venting on slashdot won't do any good.

There's a meta-implication though, which is that reinforcement learning based AI training also cannot achieve the kind of existential fear that humans get when they are threatened with consequences for serious offences.

You have too much faith in magical thinking. Filtering generative output without regard to age is an unsolvable problem for AI companies, and your naive ideas about how the law should view impersonation attempts and libel/slander (not to mention possible blackmail) are cute.

There is no world where making shit up about other people via AI is a long-term acceptable strategy.

It's meant for people with some life experience. Difficult to summarize succinctly \o/

I don't understand why these two fell for it.

They spent too much time believing the Orange Drivel machine about how America rules the World.

The issue is that actual real enforcement of laws on company executives who do bad things is no longer a thing in the American mindset. When it occurs in other countries the kneejerk response is "we don't do that, so you guys shouldn't do it either".

Two words. Look them up. 1. collage. 2. Internet.

Come on!

They're converging already!

They're converging! They're converging!

They're converging! They're converging! They're converging! They're converging!

Oooooh.

Actually, it is in most cases. Because the source lives on a server, and you had to *copy* the source into your own computer first. Then you used the software running on your computer to view your local copy.

"Mirror mirror on the wall, who is the fairest of them all?"

"You are, senator, your policies have eradicated poverty in America, next would you like to appropriate a program for sending missionaries to Mars?"

1) AI training data goes stale. Who is the president today? Etc. The value of social media training data is that it is *current*. That means when the genZ/whatever kids start saying some incomprehensible words to hide their important chats from their parents, the AI is able to know about it. Otherwise you'll soon be askinge ye olde Amazon shoppinge not for absinthe and cabbage patch dolls.

2) Meta is only the buyer of the once promising Instagram and Whatsapp brands They inherited disparate codebases that were never meant to work together, and didn't have compatible features. They are weary about making too many changes as this would break the magic, which they need (since if they could do it all themselves, then they didn't need to acquire the startups in the first place)