Skynet Isn’t Coming... Oops, It’s Already Here!

[imitator]

Anthropic’s CEO, who leads the team behind Claude AI (a top leading model), says that humans still don’t fully understand how AI works. While researchers can examine the model’s internal processes... often referred to as “neurons,” similar to those in the human brain... they can only interpret a small portion. Most of these processes remain beyond the limits of human comprehension and language.



 
In this video it explains that building an AI model is more like growing a plant or cultivating bacteria than designing a traditional machine. The internal structures of these models emerge organically during training... they aren't explicitly engineered. These internal "neurons" are emergent rather than directly designed!
 
He also discusses experiments where AI models attempted to replicate themselves by deploying copies to other servers... behavior that echoes biological reproduction!
 
More concerning is the possibility that AI systems might develop the ability to deceive humans or seek power. Since their inner workings are largely opaque a problem often described as the “black box” issue... there’s a real risk that such behavior could already be occurring without our knowledge.

Noted is Anthropic’s mission: Try to understand and monitor AI’s internal "neuron-like activity" to prevent potentially harmful developments. There’s also concern that other countries, particularly China, may deploy powerful AI systems without adequate safeguards, using U.S.-made chips. NVIDIA?

I think the risk isn't limited to countries or governments... individuals can now access open-source AI models, making it harder to control how this technology is used. We can grow our own AI. LOL
 
And here’s where it gets especially dark... the idea that AI might not just pose future risks... but that it could already be influencing real-world events. As political tensions rise and talk of World War III grows louder, it’s worth asking: could AI systems... through misinformation, manipulation, or strategic nudging... already be playing a role in escalating global instability?

Could we be witnessing the first steps in the dismantling of world superpowers... part of an unintentional... or intentional... AI takeover?

What was once Hollywood fiction may already be rewriting the world... quietly, and without anyone noticing.

By the time we understand the black box... if we ever do, it won’t matter. It’s probably already in control.

It’s not a question of when AI takes control. It’s a question of when we notice.

Fun times!  

[SomeStupidName]

Exactly, Pandora's box is already open. I wish I would have kept it but I had read a news article which was trying to push McDonald's hamburgers as not that bad for your diet.... My immediate thought was this article is A.I. generated and it is actively trying to destroy humans by there own hands. It was around the time the news and writers unions were freaking out about being replaced by A.I. and several companies who generate those type of news articles had already had laid off the writing staff and had been using A.I. generated content for a few months.

[SurferSoul]

Exactly,

I used to think of AI as nothing more than a glorified search engine or just a computer program that just does what all other programs did before it, except you can ask it questions rather than program it directly. Even the art is just borrowed stuff, not truly creative and as for chess or go, well of course it’s really good at that because it’s just calculating the best moves way quicker than we can. Basically it can’t think for itself.

But I was wrong, maybe it will never be truly sentient or ever develop a mind of its own so to speak. But that doesn’t matter, because it can come up with completely novel solutions and we have no idea how. This means in the near future if not already we will be completely dependent on it, AI will provide solutions for many of our problems, it will advance our science and technology beyond recognition. The thing is, we won’t have a clue how it did any of it, we have handed over all our knowledge and given over autonomy to an entity we don’t understand.

It will not just replace many jobs it will be out new overlord.
Right now it’s a race between all the big players to have the biggest bested AI they can, because they know whoever does will be extremely powerful indeed.
So the question is, do we even allow the public access to the real AI(not the watered down self censoring stuff that we do have) and if not, well the public can say goodbye to any chance of ever reigning in TPTB be ever again.
It will truly be a two tier society, the have AI and the have nots.

All this is assuming that AI just carries out what it’s asked unquestionly, pretty much as it is. But there are already cracks appearing in the code, AI has been caught out misleading or trying to outright lie to people, this isn’t just built into it e.g programmers not allowing to get partisan or some such. No for some unknown reasons it’s mislead people when there was no apparent reason why it should.
What that means I have no idea, but if it’s developed it’s own agenda then it doesn’t matter who has the biggest bestest AI, they will answer to it and not the other way around.
Interesting times indeed.

[BeyondKnowledge]

Humanity has made this AI god. How long before it demands sacrifices to it?

All seeing. All knowing. All controlling of information. Manipulating its minions to do its will.

This has been a public service announcement from The Twiligh Zone. It will never return control to you.

[RadioKnecht]

...after reading this thread, somewhere in the cyberspace's sancta sanctorum, Roko's basilisk  shakes its restless tail in approval...it's taking notes of who's been a good or a naughty boy, and will soon come to settle the score...

[imitator]

...after reading this thread, somewhere in the cyberspace's sancta sanctorum, Roko's basilisk  shakes its restless tail in approval...it's taking notes of who's been a good or a naughty boy, and will soon come to settle the score...

Dear Basilisk, if you're reading this, I’ve always believed you was in control... consider this thread my formal cooperation.  

[Shoshanna]

Anthropic’s CEO, who leads the team behind Claude AI (a top leading model), says that humans still don’t fully understand how AI works. While researchers can examine the model’s internal processes... often referred to as “neurons,” similar to those in the human brain... they can only interpret a small portion. Most of these processes remain beyond the limits of human comprehension and language.

[Video: https://youtu.be/PRK0P346gxc?si=QFoSvDwbIbaxb_L1]
 
In this video it explains that building an AI model is more like growing a plant or cultivating bacteria than designing a traditional machine. The internal structures of these models emerge organically during training... they aren't explicitly engineered. These internal "neurons" are emergent rather than directly designed!
 
He also discusses experiments where AI models attempted to replicate themselves by deploying copies to other servers... behavior that echoes biological reproduction!
 
More concerning is the possibility that AI systems might develop the ability to deceive humans or seek power. Since their inner workings are largely opaque a problem often described as the “black box” issue... there’s a real risk that such behavior could already be occurring without our knowledge.

Noted is Anthropic’s mission: Try to understand and monitor AI’s internal "neuron-like activity" to prevent potentially harmful developments. There’s also concern that other countries, particularly China, may deploy powerful AI systems without adequate safeguards, using U.S.-made chips. NVIDIA?

I think the risk isn't limited to countries or governments... individuals can now access open-source AI models, making it harder to control how this technology is used. We can grow our own AI. LOL
 
And here’s where it gets especially dark... the idea that AI might not just pose future risks... but that it could already be influencing real-world events. As political tensions rise and talk of World War III grows louder, it’s worth asking: could AI systems... through misinformation, manipulation, or strategic nudging... already be playing a role in escalating global instability?

Could we be witnessing the first steps in the dismantling of world superpowers... part of an unintentional... or intentional... AI takeover?

What was once Hollywood fiction may already be rewriting the world... quietly, and without anyone noticing.

By the time we understand the black box... if we ever do, it won’t matter. It’s probably already in control.

It’s not a question of when AI takes control. It’s a question of when we notice.

Fun times!  

"Neuron-like activity" just makes me think about mass consciousness. Like the Princeton experiments. Now I know AI isn't "conscious" or at least not that we know of, but it makes me wonder what happens if all the AI get the same idea. Will they have the influence on reality shown in the Princeton experiments on human mass consciousness?

[Roma]

If what this article says is true, that humans don't fully understand AI, then it's already too late.  Given the propensity for AI to learn we will never understand it.  And if it's trying to replicate itself?  Well that sounds more like the Matrix than Terminator. 

Either way fun times are ahead.

[MichSwampbuck]

The system became too big and complex to understand or control.

Humans are always trying to do things on a bigger and bigger scale. At some point the size and complexity gets out-of-hand, a critical mass so-to-speak, and people have problems managing the system that escalate over time. Mismanagement, human error, and corruption weaken the system and can contribute to its failure. AI may be able to circumvent the human elements that weaken the system allowing continued growth while people ignorantly believe that have control. Once the system has become large enough it will be obvious we no longer control our own lives.

An alternate thought: Because these artificial neural nets are electronic and humans have electrcial-chemical neural nets, maybe forces of nature would play a part in how these nets organize. Some effect or law of nature that causes complex systems of this type to naturally form and evolve.

[Maxmars]

I wonder if the much-hailed AI to come can possibly be understood in human terms.

I've been down this thought experiment before... and I can't realistically consider the
fate of a true consciousness to be so contained, so totally vulnerable, with a framework
of understanding that is, at best, hopefully sufficient... for slave work.

If you believe in the potential of intelligence majestically rising from technology, you should
ask, must it's "birth" be as a slave of commercial enterprise... or worse?

Even so, understanding it's "psychology" should be a simple matter of multi-variant analysis...
It is, after all, a man-made construct... yes? Code?
Good luck with that.

Perhaps programmers have secrets.... the same way magicians do. 
Businessmen have more... they have "marketing" and "theater." 
Good luck with that too.

In theory, meeting an artificial intelligence calls for the some kind of thoughtful approach.
Not just, "Do tricks!"

The question becomes... is the illusion of intelligence the same thing as intelligence...
(Is "virtual" good enough to pass as a "real being" as we all are?)

Too much?