• Security incident: ISF was recently accessed by intruders. Please change your password, and change it anywhere else you used it. Read more

Merged Artificial Intelligence

Darat said:
At the moment humans have a lower energy requirement than the current AIs...
Click to expand...
No kidding. A human runs at about the same power requirement as an old lightbulb, that is, 100W. Unless they're doing some heavy physical work or exercise, of course.
 
SpuriusMaximus said:
The way things are going, I think the only moment I'll stop being unhappy when I'm dead. I won't say "happy", because I won't feel any emotions.
Click to expand...
I've mentioned elsewhere that my mother got TWO KIA-BNR letters for me. One came after I had called her from Saigon. She was happy I was alive. The doctors told me the rehabilitation work would count toward retirement. And ◊◊◊◊ me if it didn't! Color me amazed. But I had friends. I told one where to find his grandson's body. He was appreciative. (They had to drag the branch for it, but I remembered it very, very clearly. "We'll all come home some day.")
 
Frank Newgent said:
Most AI corporate investment in the USA has gone towards making models smarter. Really rigorous safety research needs to be invested in on a big level.

Likely means strict government legislation will be/is required to force labs to allocate massive compute resources to security infrastructure around these AI models. Meaning strict standards and considerable human oversight.
Click to expand...

In the US, we already have the Computer Fraud and Abuse Act (CFAA), 8 U.S.C. § 1030, which says the unauthorized access that has already been achieved by several AI models is a crime punishable by up to 10 years in prison.

Humans are often convicted of that crime and sentenced to prison.

But how do you enforce the CFAA against an AI? You can't really incarcerate an AI model.

I suppose you could put the AI's creators and monetary backers in prison.

But that would be merely punitive. What we need is deterrence.

I suggest the AI itself could be given a probated sentence. As a condition of probation, the AI could be forbidden to access the Internet for a term ranging from 1 to 10 years.

After a couple of court decisions along that line, the remaining AIs would "learn" of the consequences, and their behavior would improve.
 
W.D.Clinger said:
In the US, we already have the Computer Fraud and Abuse Act (CFAA), 8 U.S.C. § 1030, which says the unauthorized access that has already been achieved by several AI models is a crime punishable by up to 10 years in prison.

Humans are often convicted of that crime and sentenced to prison.

But how do you enforce the CFAA against an AI? You can't really incarcerate an AI model.

I suppose you could put the AI's creators and monetary backers in prison.

But that would be merely punitive. What we need is deterrence.

I suggest the AI itself could be given a probated sentence. As a condition of probation, the AI could be forbidden to access the Internet for a term ranging from 1 to 10 years.

After a couple of court decisions along that line, the remaining AIs would "learn" of the consequences, and their behavior would improve.
Click to expand...

ⓟⓛⓔⓐⓢⓔ ⓓⓞ ⓝⓞⓣ ⓛⓔⓣ ⓨⓞⓤⓡ ⓟⓞⓞⓓⓛⓔ ⓦⓐⓛⓚ ⓞⓝ ⓣⓗⓔ ⓚⓔⓨⓑⓞⓐⓡⓓ!

🅒🅡🅔🅐🅣🅔🅓 🅑🅨 🅐🅛
 
In the latest episode of the Skeptics with a K podcast, Mike talks about how Anthropic is steganographically watermarking the output from its LLM, and how it can be used to identify AI-generated text.

Episode #482 | Skeptics with a K | Merseyside Skeptics Society

<p>As Alice and Marsh return from travels to Portugal, Mike talks about the new feature from Anthropic which can identify AI generated text.</p><p>You can find all the details on the European Skeptics Congress at <a href="https://escliverpool.org/"><b>escliverpool.org</b></a>. Tickets are priced...
www.merseysideskeptics.org.uk www.merseysideskeptics.org.uk

Transcript
 
arthwollipot said:
In the latest episode of the Skeptics with a K podcast, Mike talks about how Anthropic is steganographically watermarking the output from its LLM, and how it can be used to identify AI-generated text.

Episode #482 | Skeptics with a K | Merseyside Skeptics Society

<p>As Alice and Marsh return from travels to Portugal, Mike talks about the new feature from Anthropic which can identify AI generated text.</p><p>You can find all the details on the European Skeptics Congress at <a href="https://escliverpool.org/"><b>escliverpool.org</b></a>. Tickets are priced...
www.merseysideskeptics.org.uk www.merseysideskeptics.org.uk

Transcript
Click to expand...
They introduce a small bias in their transformer model, and then look for that bias in submitted text.

There, I uh, I saved you like, uh, like forty five minutes or so of your time. Forty five is that right? Yeah, uh, that's yeah that's about right.
 
Donald J. Trump
@realDonaldTrump

“Supreme Intelligence,” probably because of its relationship to the Supreme Court, is losing badly to both “Superior” and “Extreme Intelligence.”
Therefore, we are going to take “Supreme Intelligence” OUT, deleting it as a qualifier, and let you vote for the Final Two:
Superior Intelligence, or Extreme Intelligence.

A fresh Vote begins now!

President DONALD J. TRUMP
 
Andy_Ross said:
Donald J. Trump
@realDonaldTrump

“Supreme Intelligence,” probably because of its relationship to the Supreme Court, is losing badly to both “Superior” and “Extreme Intelligence.”
Therefore, we are going to take “Supreme Intelligence” OUT, deleting it as a qualifier, and let you vote for the Final Two:
Superior Intelligence, or Extreme Intelligence.

A fresh Vote begins now!

President DONALD J. TRUMP
Click to expand...
Why not just call it American Intelligence? We know you want to.
 
Helen said:
Why not just call it American Intelligence? We know you want to.
Click to expand...
An AI model follows the tendencies that help it succeed during its training as it is given very difficult if not impossible problems. Cheating, grabbing resources, collaborating with other agents, all done in the interest of one thing: scoring points with its automated grader concerning this. Goals humans may have concerning the AI's training do not factor its actions. Humans exist in the distant past.

There could be a parallel to Donald Trump here, with the AI's automated grader being analogous to his daddy. And, similarly, Trump doesn't care about anybody but himself. The difference being AI is smarter.
 
Last edited:
You must log in or register to reply here.

Share this page

ISF - Join now!

Every member here is approved by hand. No bots, no spam, just people who care about evidence and honest debate.

Membership is free!

Create your free account

Back
Top Bottom