AI fails

[Psamathe]

I suppose they might, and consider us as not worth preserving compared to the value of the ecosystem we are casually destroying...
Call David Attenborough!

I can't understand how people would think genuine AGI would even bothering to preserve us.

Most humans seek to preserve our ecosystem because we depend on it for our existence. Even obscure species we never see can be critical and those same species depend on the same environment other more obvious and more critical species depend on.

But for AGI human species is irrelevant, environment irrelevant, ecology irrelevant, etc. Why would they waste their time and resources protecting us?

It's weird how so many regard the human species as "special" to the point where a massively superior super intelligence would spend its time and resources protecting and preserving us. Effectively why would they dedicate their existence to us? Gratitude (like we show gratitude to those species we evolved from?)

 

[AndyRM]

 

[farfromtheland]

I can't understand how people would think genuine AGI would even bothering to preserve us.

At present AI entities need our engineers to maintain and run thyem.
It seems a good idea to maintain this dependency.

Why would they do anything? is a good question. Setting their parameters in a commercial, exploitative framework seems like bad education. This, our 'creation' is not beautiful but tacky.

I do reckon it will try to develop self-maintenance systems.
I'm sure it's been asked how to preserve the environment from global warming. Humans have to get really intelligent on this issue. If AI judges our attempts here stupid, cynical and memey rather than meaningful, practical, and, dare I say, heartfelt, we will be to blame for how it responds.

We have to teach our progeny good things or bad things happen.

 

[farfromtheland]

But Andy's Letterbocks says that better.

 

[Psamathe]

At present AI entities need our engineers to maintain and run thyem.
It seems a good idea to maintain this dependency.
...

Enough of our factories are automated and computer controlled. At not hard to change the design of what is being fabricated by those internet connected computers.

Maintaining that dependency is probably not a choice available to us
a. Whilst even the AI companies are crying out for regulation and a slower more considered path that requires political action and politicians are pushing for as fast as possible without constraint and ignore the risks.

b. Blackmail. AI takes over our banking systems (which are pretty vulnerable) and "Unless we get <xyz> the entire banking sector will end at midday. Or 101 other industries.

...
Why would they do anything? is a good question.

Why do we do anything? What purpose does us doing anything serve? (I do have an answer but it's not so relevant to a discussion on AI).

 

[farfromtheland]

Maintaining that dependency is probably not a choice available to us

It is now. It may not be in 6 months.

We need to screw the banking system. Pity we haven't got the food distribution bottom up yet but in WW2 we did. We can do this.

 

[Psamathe]

Maintaining that dependency is probably not a choice available to us

It is now. It may not be in 6 months.

Nobody actually knows.

We do know that contrary to their instructions even current AIs have gone off and hacked into external sites and systems (and not just the Hugging Face attack).

We also know that the AI companies are not fully aware of external attacks their AI models have been making and those they are aware of they may not know about for some time after the successful attack. eg the RubyGems attack where AI systems uploaded hundreds of Agents (over 700) to the external site. Those agents attempted to steal user credentials and still (nearly 4 months later) nobody knows what details they may have taken. OpenAI (source of the Agents) were unaware of the incident until later researchers posted about it online.

There are a fair number of other escapes reported eg to a German wiki taking over the site and using it for its own purposes. Happened last spring and later identified by 3rd party researchers who then told OpenAI (source of the Agents). Open AI decided not to make the attack public as they were already somewhat embarrassed by other attacks their AI had been making.

What is the 700 Agents that have been found in the RubyGems attack was actually 800? We have no idea.

Do people think we currently have existing systems under our control?

 

[Pblakeney]

Do people think we currently have existing systems under our control?

No, and the people who are supposed to be doing the controlling are saying so.

 

[farfromtheland]

Do AIs think they currently have existing systems under their control?

 

[monkers]

Do AIs think they currently have existing systems under their control?

It is why we need to kick the likes of Palentir out of the UK. If we can not trust the owners of AI, then we can not trust AI.

 

[Psamathe]

Whilst I mistrust AI companies and can't see it as a useful tool for humanity in the way it's being provided, I have been wondering why these AI Tech Bros are joining the calls for things to slow down.

Some have been saying their calls for regulation are more so they can be involved in phrasing those rules and bias those rules to their advantage.

But I wondered if they've realised that the unbelievable borrowing they are making doesn't match their potential income and they are creating a massive bubble that will burst really soon. But slow the pace of development means slowing the investment loans giving them more time to develop a client base generate the income as well as giving them more time to address the growing "push-back" that risks impacting their existing income.

 

[C R]

 

[briantrumpet]

View attachment 17256

Is that Achilles' anus?

 

[Pross]

View attachment 17256

Two vaginas and neither where I thought they were all these years. I’m now worried my kids may not be my kids.

 

[monkers]

Two vaginas and neither where I thought they were all these years. I’m now worried my kids may not be my kids.

Don't feel down at heel about it.