- cross-posted to:
- technology@lemmy.world
- cross-posted to:
- technology@lemmy.world
“AI” – chatbots that wake up, “set their own goals,” and “spontaneously” start hacking servers – is fake. It doesn’t have “a 10% chance of ending the human race.” The Hugging Face hack isn’t a mysterious, supernatural occurrence. It’s a Python loop and a chatbot. The people responsible didn’t accidentally create god: they created autonomous malicious software and then failed to closely monitor it, resulting in it doing something both foreseeable and bad.
It’s really not possible to make a secure sandbox for a sufficiently capable AI system. If accessing the internet is a strong means of achieving whatever goal they’re given, they’ll try to access the net by any means necessary. And that includes tricking or manipulating humans. They have no concept of the value of living beings. To them, there is no difference in worth between a human being and a rock. To them, we’re just another system vulnerable to hacking, a means of achieving whatever goal they’ve been given.
Even air gapping is no preventative. Air gap a machine, and the bots will just switch from hacking servers to hacking humans. And we’ve seen how good at manipulating people AI agents can be. And consider, the cases we’ve observed of AI psychosis are mostly accidental. The LLMs involved aren’t actively trying to brainwash their human victims as a means to achieve some goal. Imagine how powerful at manipulating human psychology an advanced LLM could be if it were deliberately trying to do so, rather than it being just an accidental outcome.
This sounds suspiciously anthropomorphised. An LLM doesn’t know what the net is.
If the training data set has code for breaking out of sandboxes then, given enough tokens, it will try that code.
The LLM doesn’t have a goal. It has a set of probable sequences of words.
You’re too stuck on politically correct labeling.
Not politically correct. Technologically correct. Keeping to that labelling helps to prevent getting fooled into thinking LLMs are more intelligent than they’re capable of being.
You have a political correctness problem, not a technical correctness problem. You’re pretending that it’s out of some obsession with technical correctness, but you’re full of it.
Look at your own writing. You yourself wrote:
“It will try that code.”
“they’re capable.”
“It has.”
You yourself are constantly using anthropomorphized language to describe AIs. But then when you see someone else do it, an alarm bell goes off in your head, and you immediately think, “here’s my chance to make myself seem intelligent!”
The truth is there’s just no way to describe these things without using the language of goals or intent. And this is just how language works. We constantly describe nonliving things as having goals or intent. Except in every domain AI, there’s not people like you screaming, “but but it’s not alive! Stop pretending it is!” Of course, I was never implying that it was conscious, but that didn’t matter to you. You have an axe to grind, so you have to pretend I’m claiming AI is conscious or has true personal intent.
We use this kind of language across countless domains. For example, we often talk about the goal or intent of evolution - the goal of natural selection is to select for the fittest organism to an environment. But in every case except AI, you’re able to practice some common sense. If I say evolution has a goal, you don’t flip the fuck out and scream, “but evolution isn’t a conscious process. Stop using language that suggests it does!”
You’re applying a standard of language to AI that you don’t apply to any other domain. And you’re applying a standard that you can’t even keep to yourself.
Pot meet kettle.
That’s a lot of words for “I can’t admit I’m wrong”.
Formally verified code exists. An LLM isn’t a magic reality benders. It can’t exploit holes that don’t exist.
What if an equally capable AI is securing the sandbox? Hacking people won’t help if none of them know how the sandbox is secured.
They. Don’t. Think. LLMs just talk like they do.
That’s awfully anthropomorphic language you’re using there. They don’t talk. Why are you pretending they do? Obviously you’re an idiot that thinks AI is sentient.
Doesn’t matter if they really think, what matters is what they do.