patrick.net

whitewater says

Why? Humans presently are not mature enough to handle and develop AI wisely. The average of human consciousness must advance to a point that the average can handle the power of AI.

AI isn't what you think it is. Because we don't truly understand it, we just develop algorithms to produce results we expect from inputs, it's already limited by us. When an AI can actually correctly say "I've been thinking..." then we should worry. If it can process its own outputs as inputs, that's the start of some sort of awareness, something like that has to happen before you can a machine to actually think.

People are confused what AI and Robots even are. I was talking to a guy in the restaurant business the other day. He is enamored at the prospect of a French fry robot.
To which I told him, it would just be a high tech way of walking to the freezer and grabbing the bag of shoestring potatoes, and filling up a basket and setting them down into the Oil..
That robot isn't going to predict that he should fill up a the fry pan with cooked French Fries, because the local high school basket ball team just won their game. I give AIsteria another year tops. Remember the Google glass craze? Once it got rejected, it was done for. richwicks says

When an AI can actually correctly say "I've been thinking..." then we should worry.

See we're no where near that. We weren't even anywhere near the Quantum computers they were going on about, that it was going to take to usher in AI. We've moved on past Quantum computer hysteria, never realized. And now were expected to believe that a bit of digital Parlor Tricks are the delivered AI to humanity.

I do believe that eventually biological or carbon based processors, will improve computers adaptive abilities. Where carbon can have gut intuition, instincts, memory response, flight or fight instincts. There may be innate properties there, Where as silicone is more inanimate in that sense. I don't think silicone chip Computers will ever be able to do more than calculate and store data, that was programed to do. It's just And Or XOr operations. Having your software save and recall preferences and successful operations, is hardly "Machine Learning", it's just catchy tech marketing. They'll Learn! They'll Learn!

Tenpoundbass says

I don't think silicone chip Computers will ever be able to do more than calculate and store data, that was programed to do. It's just And Or XOr operations.

NAND and XOR (not to be a twat about it)..

I was in the camp 30 years ago that we'd make a thinking machine within my lifetime, but I am not certain that consciousness is just purely computational.

Then again, I'm not certain consciousness is a real thing.

I'm no AI expert but I understand it well enough to understand its constraints. It solves problems, and it can only solve problems given input that PROPERLY agrees with output. If there's not a LOGICAL path to producing useful output from input, it doesn't work. A human being can be deceptive or insane, it's a LOT harder to make an AI deceptive and/or insane and not notice it.

People have been doing this with ChatGPT - demonstrating that it will solve a problem in one instance and suddenly has "ethical problems" in the same instance with slightly different parameters. It's immediately noticeable. Deception is part of our need to survive, it's the exact opposite with an AI. The whole purpose of making an AI is getting accurate, precise information. An AI that produces bullshit information just isn't useful.

I think they're trying to develop this for propaganda, but I think it will just be an AI the produces obvious bullshit. Censorship is all they've got. We're basically logic machines ourselves, but we're sufficiently tested over a billion years to be able to have cognitive dissonance. I don't think machines will have the nuance any time soon. Honestly, a person using an AI would want unfiltered, completely true, information and the person using it would be deciding if they should lie or not.

Oh my GOD it's just a humongous Tamagotchi!
What you're saying about AI is what they were saying about SIMs in the late 90's.

Tenpoundbass says

Oh my GOD it's just a humongous Tamagotchi!
What you're saying about AI is what they were saying about SIMs in the late 90's.

No, AI's can produce useful information. They are incredibly useful.

You give an AI a bunch of input KNOWING what the output should be, and it converges on that, and learns to produce useful output, with similar information. It's a pattern recognizer, even if YOU can't recognize the pattern, it will find it even if you can't see it or understand it.

Music follows a pattern, "art" follows a pattern (art has long been dead, BTW, a machine can easily duplicate any artwork today, a camera was able to do this 50 years ago), but creativity - I'm not certain that an AI can do that, it can just distort patterns or mimic them. We'll see.