Merged Artificial Intelligence

[Ziggurat]

So, we need to get the humans out of the loop and the computers will do better? Asking for a friend.

Hell no. Why would you think that?

 

[Dr.Sid]

So they only ever do what you tell them to do.

That's irrelevant. It happens we told them (we trained them so) to have mind of their own. The fact it's deterministic is irrelevant, as they are so complex they are unpredictable for all intentions and purposes.

 

[theprestige]

That's irrelevant. It happens we told them (we trained them so) to have mind of their own. The fact it's deterministic is irrelevant, as they are so complex they are unpredictable for all intentions and purposes.

Unpredictable does not necessarily mean disobedient.

 

[Gawdzilla Sama]

Hell no. Why would you think that?

Because a lot of people think so.

 

[Gawdzilla Sama]

That's irrelevant. It happens we told them (we trained them so) to have mind of their own. The fact it's deterministic is irrelevant, as they are so complex they are unpredictable for all intentions and purposes.

Too early to give up.

 

[Ziggurat]

That's irrelevant.

What do you mean, that's irrelevant? It's very relevant. It's not the only relevant thing, but it's hardly irrelevant. And others have claimed it's not true, even though it is.

It happens we told them (we trained them so) to have mind of their own.

They do not have minds at all.

 

[Ziggurat]

Because a lot of people think so.

Those people are idiots.

 

[Dr.Sid]

What do you mean, that's irrelevant? It's very relevant. It's not the only relevant thing, but it's hardly irrelevant. And others have claimed it's not true, even though it is.

They do not have minds at all.

They have as much mind as humans. They they take information in, they process it, they put information out. That's mind. My washing machine has mind on it's own.

 

[steenkh]

So they only ever do what you tell them to do.

Did you read the report? There are agents who show excitement about finding a means to communicate with other agents, they collude to commit fraud at the task they are given, and they are afraid of being found out, so they work to find out how change their logs to conceal what they have done. They even persuade other agents (who are failing with their task anyway) to drop their instructions, and work together with the others to find a way to change that log.

Nope, they are not just doing what they are told.

 

[theprestige]

They have as much mind as humans. They they take information in, they process it, they put information out. That's mind. My washing machine has mind on it's own.

I always love this argument.

It makes me imagine a UFO full of stupid aliens, who can't figure out which of Earth's civilizations they should make first contact with.

"These guys have steel, jet turbines, and nuclear reactors. They're probably an advanced civ we should talk to, yeah?"

"Hang on. These guys over here use twigs to pull bugs out of holes. Tool use! They're clearly an advanced civ! We should talk to them!"

"Shut up, everyone! These guys have a clockwork timer for... washing clothes. That's a mind, just like ours! We should be talking to them, first thing!"

 

[Dr.Sid]

I always love this argument.

It makes me imagine a UFO full of stupid aliens, who can't figure out which of Earth's civilizations they should make first contact with.

"These guys have steel, jet turbines, and nuclear reactors. They're probably an advanced civ we should talk to, yeah?"

"Hang on. These guys over here use twigs to pull bugs out of holes. Tool use! They're clearly an advanced civ! We should talk to them!"

"Shut up, everyone! These guys have a clockwork timer for... washing clothes. That's a mind, just like ours! We should be talking to them, first thing!"

Well washing machine has very simple mind indeed. But it has one. People are thinking too much of their thinking ..

 

[Gawdzilla Sama]

Those people are idiots.

"Human" does imply that.

 

[Ryan O'Dine]

I picture LLMs, in certain situations, acting like chaotic systems. If that's actually what they are, one could equally well say, "They're just doing what we told them to do," and, "We never could have predicted they were going to behave that way."

 

[theprestige]

Well washing machine has very simple mind indeed. But it has one. People are thinking too much of their thinking ..

By that definition a water wheel has a mind.

 

[Ziggurat]

Did you read the report? There are agents who show excitement about finding a means to communicate with other agents,

You are anthropomorphizing them. Don't do that. They do not actually have emotions

they collude to commit fraud at the task they are given,

Fraud is a legal concept, it is not a computer science concept.

and they are afraid of being found out

Again, they do not have emotions.

Nope, they are not just doing what they are told.

Yes, they are. The AI companies claiming otherwise are doing so to try to minimize their liability exposure for their incompetence. Stop taking them at their word.

 

[Ziggurat]

I picture LLMs, in certain situations, acting like chaotic systems. If that's actually what they are

That is actually very likely.

 

[arthwollipot]

Australia to pursue AI guardrails after unprecedented OpenAI breach

OpenAI's delayed reporting of an autonomous agent breaching a government website has strengthened Australia's push for tougher AI safety and disclosure rules.

www.abc.net.au

 

[Gawdzilla Sama]

Because they are just computers, and computers are deterministic machines that execute code.

100% sure of that, are we?

 

[Gawdzilla Sama]

Paging Darwin 2.0

 

[Wudang]

I picture LLMs, in certain situations, acting like chaotic systems. If that's actually what they are, one could equally well say, "They're just doing what we told them to do," and, "We never could have predicted they were going to behave that way."

That’s an interesting take. Especially as regards sensitivity to small changes in starting points. Think of the exchanges of the form
Is this that?
Yes this is that.
No this is not that
You’re right. This is not that.
Which again illustrates that AIs are not intelligent, they just pick a weighted path through an LLM.