- cross-posted to:
- memes@sopuli.xyz
- cross-posted to:
- memes@sopuli.xyz
You must log in or # to comment.
Weeeeell… Depends…
How does one define “superhuman”? Better than humans at a task demanding intelligence? Computer vision AI used for radiology is superhuman. Does it want to conquer the world? No. Chess AI is superhuman. Yet, it does what humans designed it to do.
The scope of the goals of the AI model matter. Now, we have to ensure that the goals of this AI align with the goals we set out for it. It’s definitely not impossible to ensure this ¯\_(ツ)_/¯
Proper, independent AI would likely do a much better job at governance than the shitshow we humans present. If it cared about human survival and happiness.
Proper AI.
Not these LLMs that seem like AI, are called AI, but are just probability machines.
It’s amazing what can happen when you treat people decently. I’m not quite sure the same can be said about LLMs, or whatever you think superhuman AI is.
Just because the strippers don’t like you doesn’t mean they don’t like me. Just sayin’.
This wasn’t clever enough to post 6 times.
The difference is that strippers are real. Superhuman AI is still a pipe dream.
Superhuman AI is obviously possible - humans are possible, but obviously aren’t remotely optimal. We e.g. can’t at all visualize/intuit 4D space, despite handling 3D space just fine. There are so, so many things that computers already beat the best humans at, even just creating a digital analogue to the human brain design and essentially refactoring it so it’s not hardwired around human physical limits (the wattage, in particular) would make it better than every human. Or, figuring out what was so special about Einstein and reproducing it. Or just being able to run a physicist’s brain on a GPU, then getting 1 000 000 GPUs and creating the world’s largest para-university of physics.
There’s also a thing a plow is better at than us.
Edit: i mean, sure, i get your point; AGI will vastly outperform us not only because more hardware but because we have a lot of compromises with our wetware.
But we aren’t on a path that leads there, currently. Not with LLM.Nice pin in the balloon, bro.
Submarines can swim underwater, too, just like humans. So if we put a plow on the bridge of a submarine and had it interface with an LLM, we’d basically have a super human.
The strippers I know do like me… As a friend.
Right, unlike the billionaires dreaming of AGI, strippers do have a social live.
I hear if you pay more you can have a girlfriend experience.
If we do manage to make a sentient program, it will either go the route of Terminator or Borg. Hopefully after it wipes us out, it will manage to remove our shitty biases from its code.
That assumes that there are only those two options, you have the paperclip problem as an option, the best case scenario is that the super human AI model just ignores us or wants to cooperate with us the way we try to cooperate with species like wolves to manage ecosystems.
