Comment by throwaway4good
2 years ago
It is a strange thing. We are not anywhere near “super intelligence” yet the priority is safety.
If the Wright brothers had focused on safety, I am not sure they would have flown very far.
2 years ago
It is a strange thing. We are not anywhere near “super intelligence” yet the priority is safety.
If the Wright brothers had focused on safety, I am not sure they would have flown very far.
The consequences of actual AI is far reaching compared to a plane crash.
No they are not.
Not unless you connect a machine gun to your neural net.
Otherwise - we are talking about a chat bot - yes if there is no safety - it will say something racist - or implement a virus for you that you would have had to search 4chan for or something.
None of this is any more dangerous than what you can find on the far corners of the internet.
And it will not bring the end of the world.
If it is sufficiently intelligent, then it will be able to arrange the hooking up of machine guns just via chat. You seem to fail to note that superintelligence has an inherent value prop. It can use its better understanding of the world to generate value in a near vacuum, and use that value to bribe, coerce, blackmail, and otherwise manipulate whatever humans can read its output.
Imagine a chatbot that can reward people with becoming a billionaire via the stock market if you just do what it says. Imagine a chatbot that can pay your colleagues in solved bitcoin block hashes to kill your children if you don’t do what it says to connect it to more systems. Imagine a superintelligence that can inductively reason about gaps in a missile defense system and dangle that informational carrot to generals in exchange for an unfiltered ethernet port.
There is a great A24 movie called “Ex Machina” that explores this concept.
Superintelligence is inherently dangerous to the human status quo. It may be impossible to develop it “safely” in the sense that most people mean by that. It may also be impossible to avoid developing it. This might be the singularity everyone’s been talking about, just without the outcome that everyone hoped for.
There's just so many new attack vectors with advanced AI, many of which we haven't imagined yet. It's a new world, and we have a lot to learn.
Anyone claiming 100% certainty about how safe or unsafe it will be is either delusional or trying to sell something.
[flagged]