this post was submitted on 30 Oct 2023
546 points (94.8% liked)
Technology
59207 readers
3110 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
It won't end the world because AI doesn't work the way that Hollywood portrays it.
No AI has ever been shown to have self agency, if it's not given instructions it'll just sit there. Even a human child would attempt to leave room if left alone in there.
So the real risk is not that and AI will decide to destroy humanity it's that a human will tell the AI to destroy their enemies.
But then you just get back around to mutually assured destruction, if you tell your self redesigning thinking weapon to attack me I'll tell my self redesigning thinking weapon to attack you.
I’m an AI researcher at one of the world’s top universities on the topic. While you are correct that no AI has demonstrated self-agency, it doesn’t mean that it won’t imitate such actions.
These days, when people think AI, they mostly are referring to Language Models as these are what most people will interact with. A language model is trained on a corpus of documents. In the event of Large Language Models like ChatGPT, they are trained on just about any written document in existence. This includes Hollywood scripts and short stories concerning sentient AI.
If put in the right starting conditions by a user, any language model will start to behave as if it were sentient, imitating the training data from its corpus. This could have serious consequences if not protected against.
There are already instances where chat bots demonstrated unintended racism. The monumental goal of creating a general purpose intelligence is now plausible. The hardware has caught up with the ambitions of decades past. Maybe ChatGPT's model has no real hope for sentience, as it's just a word factory, other approaches might. Spiking neural networks for example, on a massive scale, might simulate the human brain to where the network actually ponders its existence.
AI does, but we haven't developed AI and have no idea how to. The thing everyone calls AI today is just really good ML.
At some point ML (machine learning) becomes undistinguishable from BL (biological learning).
Whether there is any actual "intelligence" involved in either, hasn't been proven yet.
Imagine 9-11 with prions. MAD depends on everyone being rational self-interested without a very alien value system. It really only works in the case you got like three governments pointing nukes at each other. It doesn't work if the group doesn't care about tomorrow or thinks that they are going into heaven or is convinced that they can't be killed or any other of the deranged reasons that motivate people to do these types of acts.
The real risk is that humans will use AIs to asses the risk/benefits of starting a war... and an AI will give them the "go ahead" without considering mutually assured destruction from everyone else doing exactly the same.
It's not that AIs will get super-human, it's that humans will blindly trust limited AIs and exterminate each other.