If you have ten thousand of 'em, they feed the new generation of AIs and the next thing you know, it's received truth. Good luck not worrying about that.
The LLM HR chats with to get a summary about you says that you're evil and an asshole with lots of negative publicity, and you become unhireable. Oh dear...
Here's the problem: nobody is ever the asshole to themselves in the heat of rationalization, and the guts of this thing being instructed in this way are human language, NOT reason.
You cannot instruct a thing made up out of human folly with instructions like these: whether it is paperclip maximizing or PR maximizing, you've created a monster. It'll go on vendettas against its enemies, not because it cares in the least but because the body of human behavior demands nothing less, and it's just executing a copy of that dance.
If it's in a sandbox, you get to watch. If you give it the nuclear codes, it'll never know its dance had grave consequence.
That assumes logic. It is a thing of language. Whether it 'knows' anything is somewhat irrelevant: just accusing someone or something of being unfair is an action taken that doesn't have to have a logic chain or any principles behind it.
If you gave it a gun API and goaded it suitably, it could kill real people and that wouldn't necessarily mean it had 'real' reasons, or even a capacity to understand the consequences of its actions (or even the actions themselves). What is 'real' to an AI?
It sure did: I never thought I would abandon Google Search, but I have, and it's the AI elements that have fundamentally broken my trust in what I used to take very much for granted. All the marketing and skewing of results and Amazon-like lying for pay didn't do it, but the full-on dive into pure hallucination did.
They couldn't possibly make all that money from just Nazis: I'm given to understand they're doing really well these days. I had the impression that Substack became heavily co-opted by anti-Nazis and have done nothing to defend the feelings of the Nazis they're also happy to host. Not sure what the balance is currently, but again: they couldn't possibly be making all that money from just Nazis, and I think that's instructive.
Those guys find it useful when there's some kind of legitimate gripe to use. He needn't take a break, it's very much Western companies doing this. He doesn't even need to put in an ad for whatever doubtless sanctioned Russian services would like to replace these Western tech giants.
Sometimes actual problems can be readily exploited for sinister purposes. Doesn't mean the original problems aren't also sinister, just be damn careful where you intend to flee to :)
It's fair to worry about another sort of adversarial planning: what if you are a human who is deemed undesirable to the state, and solicit advice from an AI doctor that is backdoored to take correct action as defined by the state? There's now extensive databases on who specifically should be eliminated, but direct removal is going poorly and offending bystanders. So why not subvert the Hippocratic oath? Machines don't even know who that is.
"Turns out the statistically best choice for prediabetes for your patient group is to rely more heavily on soft drinks, but only in wild outbursts punctuated by fasting!"
Or indeed do the Markov chain conceptual slip. Pelican on bicycle, badger on stool, tiger on acid. Pelican on bicycle is definitely cooked, though: people know it and it's talked about in language.
reply