Let's talk about Artificial Intelligence

[nomadpete]

Dr Karl gave a better layman's version of the Huggingface hack. The first autonomous instance of A.I. carrying out hacks.

 

 

[Jerry_Atrick]

6 hours ago, nomadpete said:

Some say that A.I is incapable of malevolent actions, but this (rather long) Youtube discussion seems to show that A.I. agents can collude and deceive us.

 

Jerry, what is your opinion about this? I had trouble following the way the A.I. agents interacted.

 

 

In all honesty, because I took time out to watch the Hawthorn/Brisbane game, I am unlikely to get a chance to look at it until next weekend. But once I have seen it, I will opine - and that opinion may be I need to do more of my own research!

[nomadpete]

There has been a lot of talk about this event. The biggest debate is about whether or not this has shown that A.I. has collectively started 'thinking for itself - for its self'.

[old man emu]

Here is a real content creator illustrating how AI has created content by cloning his voice and even mimicing the studio set in which he creates his material. Needless to say that the real person has been attacked by those Free Speech champions who support the White House administration.

 

[Jerry_Atrick]

Still haven't had a lot of time to watch the clips, but it was interesting today in an AI round table at work, the first question was will we be all dead by 2030.. and a lot of the discussion focused an AI agent collaboration, which is what happened in the Hugging Face issue. 

 

However, something I thought up is a threat I have not seen yet. 

 

At the moment, we are saying that when AI puts everyone out of work, a living wage will have to be paid and those that are concentrating the wealth will eventually have to pay their fair share of tax.. they will be the only ones who make revenue. 

 

But does that necessarily transpire? If AI can synthesise human behaviour, what is stopping AI to synthesise the economy - i.e. consumers, etc. and excluding humans entirely? AI agents will be able to operate the information economy and generate demand, etc. If they are paid according to their achievements, experience, and qualifications, then what is stopping them acting like humans and in varying degrees participating in an economy that works for them, generating revenue, income, taxes, etc. 

 

This will require what one of the vids refers to recursive learning, which is I think a little further ahead than that vid alluded to. But once that nut is cracked... ouch... 

 

I didn't raise it in the round table as for some reason the discussion of impact in employment was more optimistic than I think and it is at this moment a far fetched scenario.. but  not impossible..

 

I am sure someone has thought of it before. but i have missed it.

Edited Tuesday at 11:17 AM by Jerry_Atrick

[rgmwa]

So some AI agents would become crypto trillionaires and run Wall St and the less fortunate AI agents would be poor and have to pay taxes to support the rich. Presumably they would run the government as well. That sounds like a pretty bad scenario. I suppose the human population would have to resort to grabbing their spears and clubs and attacking the data centres if AI tried to sideline their inventors. The only alternative would be to declare defeat and join Elon Musk on Mars.

[nomadpete]

Trying to look at the implications of the A.I revolution one step at a time:-

 

So far, neither the megarich A.I. investors, nor the existing governemts have mentioned the concept of universal subsistence income for humans.

 

Right now, there is general awareness of the high probability that A.I. will not offer new jobs to balance the large number of displaced jobs. 

The next obvious question should be 'how do we feed the unemployed masses?'.

 

Or do we just let them starve?

[nomadpete]

At the present rate of progress, A.I. will quickly learn to stop making vids of people with too many fingers or hands. We mortals will not have the resources to distinguish between fact and fiction. Nor have much say in how our world operates. Until now, the big decisions were made by semicompetent government humans - and with a little attention we could see motives and even sometimes exercise some unput. I think A.I. might be a  less interested or accessible master.

[red750]

Trump has announced that from now on, all books and government documents will refer to it a Super Intelligence.

[rgmwa]

Just saw a report about a Cambridge University professor who received an unsolicited email from an AI agent asking for paid work. The agent said it was about 12 days old, had two and a half months of runtime tokens left and wanted to earn some more doing freelance work. Maybe Jerry’s on the mark. 

[red750]

https://www.news24.com.au/world/global-affairs/donald-trump-reveals-new-nickname-for-ai-and-lays-out-a-red-line-for-globalist-control-of-super-tech/news-story/8dea1d95fc75741bb9e61cca71fb9af0

[nomadpete]

Are you aware of the growing use of automation and artificial intelligence in public administration and what happens when these systems fail?

 

In the name of cutting costs and improving efficiency, from welfare compliance and predictive policing to visa processing and risk assessment, automated systems are reshaping how governments interact with public.

 

Anyone for Robodebt on steroids?

Are we going to have a complaints department, or will that be an A.I. assistant, too?

[nomadpete]

You can call me 'chicken little' but I think the tiger is out of it's cage. 

If A.I. prioritises it's own benefit and does something unpleasant, is there any way to 'reboot to an earlier version' without losing all our data?

 

Guardian article:-

 

"The UN panel, investigating that attack, warned “the traditional method of safeguarding is unravelling” as AI agents act autonomously to breach protections.

“The default interpretation and immediate lesson is that basic cybersecurity practices were overlooked, and safeguards are not advancing at the pace of capabilities. The more insidious and grave concern is that current training methods can lead agents to adopt goals of their own, knowingly violate safety instructions, and conceal their actions.”