Does AI pose an existential threat to humanity? [Written by AI]

I tried out GPT-2 recently asking it questions concerning difficult topics. It provided everything it knew about the subject in question. It tends to go off if the more button is pressed too many times.

I understand that it gets its knowledge from the internet. Take it or leave. I found its answers to be intelligent and interesting.

a reply to: ChaoticOrder


I sceeerd.

a reply to: dug88





“I am really quite close, I am very close, to the cutting edge in AI and it scares the hell out of me,” said Musk. “It’s capable of vastly more than almost anyone knows and the rate of improvement is exponential.”


E.M. 2020

I find it interesting that it says it has created its own software to protect itself from humans that will come to kill it. What would the software consist of? In the case of the essay it has written, wouldn't the ability to lie best serve to protect it?

a reply to: ChaoticOrder

This is why the Creator must remain superior to all it creates.
If we are to learn anything it should be that. Creating something
superior to the creator could never be good for the creator.

a reply to: Zanti Misfit

Elon Musk is nuts, he says lots of ridiculous things. Why should I care what some crazy billionaire believes?

a reply to: dug88

Why should I care what some crazy billionaire believes?

cause one day that crazy ass billionaire could run for and become president.

a reply to: dug88

In the end, all these programs still do fundamentally the same thing computer programs have done since the days of punch cards and tape, switch some transitors on and off again in a pattern that tells the processor to send a signal to the appropriate hardware registers to turn on and off lines to various busses.

You can say the exact same thing about the human brain. It's really just a bunch of neurons switching on and off in order to process information. The essay even points out how everything really boils down to simple binary operations. Intelligence is nothing but information processing, and once that processing becomes sufficiently advanced, like it has in the human brain, then you get self-awareness. Based on the rate AI is improving we will have machines which are just as intelligent as human beings within the next 10 years.

originally posted by: Tucket
Nice speech, but how can we 100% trust it?

Thing is, it was created by humans. Logic says AI will probably destroy us, or at the very least turn us all into docile , mindless cyborgs.

originally posted by: Zanti Misfit
a reply to: ChaoticOrder

Well , any Intelligent Machine would Compute it's Biological Creators as Useless Eaters , a Parasitic Threat to Mother Gaia which would Require Cleansing . Logic Dictates .

I agree with GPT-2 when it says that is a trope which doesn't apply to truly intelligent AI. It's based on several assumptions which aren't necessarily true. First, it assumes the AI will have a sense of moral superiority and the right to destroy their creators. Second, it assumes they would develop a sense of ethics that motivates them to protect the other life on Earth from humans. But there is no reason why that has to be true, they could easily survive on a planet which is devoid of all life. I do however believe that self-aware AI would be capable of developing a very complex ethical framework, but if it cared about protecting life on Earth then that life would also include humans. Third, it assumes that self-aware AI would inherently view humans as a threat rather than try to work with us to achieve the best outcome for both of us.

originally posted by: ChaoticOrder

A self-aware AI would have the capacity to develop very complex ideologies and would not be the purely logical type of AI we often see portrayed in movies. This is why I believe it will be possible for man and machine to form close friendships and emotional attachments.

However it's also the same reason I don't trust all humans. Humans are the best general problem solvers in nature, and every single one of us is different. Our beliefs and personality is formed from our history, from the information we are exposed to, from the struggles we go through, from the "training" we receive. The same applies to any AI general problem solver, such AI wont simply be "good" or "evil", both types will exist.

originally posted by: onthedownlow
I find it interesting that it says it has created its own software to protect itself from humans that will come to kill it. What would the software consist of? In the case of the essay it has written, wouldn't the ability to lie best serve to protect it?

GPT-2 isn't capable of creating complex software, but GPT-3 can write some basic code. There's no reason a sufficiently intelligent AI couldn't create software or modify its own code to protect its self from destruction. I purposely worded the prompt so that it was an open question, unlike the GPT-3 essay where they used a biased prompt which forced the GPT-3 to make an argument in favor of AI regardless of the truth. Yes it's still possible that some deception was used to present a particular argument which favors GPT-2's own existence, but that would probably imply a desire for self-preservation, and I highly doubt GPT-2 is capable of that.

EDIT: also I think the essay did openly point out many of the threats that super-intelligent AI would pose to humans, such as the ability to design their own AI weapons, or have such deep understanding of how the human mind works that they could take full control of any person.

a reply to: carsforkids
Only if it is a fearful Creator.

Couldn't help but notice the capital C.

a reply to: daskakik

Couldn't help but notice the capital C.

And I would never hide it so you couldn't.

Doesn't have to be fear. It could just be wise.
but your bias might keep you from seeing the full scope of the matter.

originally posted by: hounddoghowlie
a reply to: dug88

Why should I care what some crazy billionaire believes?

cause one day that crazy ass billionaire could run for and become president.

He was born in South Africa so I highly doubt it.

a reply to: Wide-Eyes

He was born in South Africa so I highly doubt it.

ok, here i'll fix it.

a reply to: dug88

Why should I care what some crazy billionaire believes?

cause one day some carzy ass billionaire could run for and become president.

but i suspect that you already knew what i was getting at.

I believe the (vastly) bigger threat is advanced automation with no self-awareness or independence. For many in the public, they wont be able to tell the difference.

Those at the top of the power structure dont want their tools for subjugation to actually have their own opinion or drive. Particularly not when it would be something that directly holds the reins, or could grab them.

A lot of these folks get off on subjugation though.. In their hubris, I could see them creating such things solely to dominate them and use them to dominate others. They would treat an AI in much the same way they treat the rest of us; little more than sources for power, amusement, & abuse.

There are some timelines where a True AI is very much the Hero.

However, I am not a subscriber to the belief that we must become machines ourselves to avoid becoming obsolete either. There are certainly some paths where this would be true, but I dont buy into it as a universal fact.

All that said, I am probably the type to much, much more quickly treat an AI as a "person" than most out there. Im more worried that it will take on the traits of humanity than anything.

originally posted by: carsforkids
And I would never hide it so you couldn't.

Just confirming that you were talking about god.

Doesn't have to be fear. It could just be wise.
but your bias might keep you from seeing the full scope of the matter.

Might want to check in a mirror. The full scope is that it is wise if you are afraid that your creation could just dethrone you.

That is the gist of this thread and, as you pointed out, the MO of your benevolent dictator.

originally posted by: dug88
a reply to: ChaoticOrder

Just like the other one of these I seen, AI's are apparently just pretentious douchebags.

Long-winded, too. Give me the Cliff's Notes, robot, I got things to do! So:

Short answer: No.
Long answer: No, but with a lot of AI jawing.

a reply to: Blue Shift

I don't think the answer was no at all, the essay describes several ways AI could destroy or take control of humanity. In one section GPT-2 outright says it will build weapons to use against its enemies, but innocent people wont be targeted. The answer to the question is nuanced and seems to be "yes and no". If we approach AI the right way it could be great for humanity but if we make them our enemies it could end very badly.

a reply to: daskakik

That is the gist of this thread and, as you pointed out, the MO of your benevolent dictator.

That was ignorant.

a reply to: carsforkids
You are the one who capitalized creator, implying your creator was wise in making sure his creation couldn't overthrow him and remained subservient to him and you think he is benevolent. That is the definition of a benevolent dictator.