Anthropic warnt davor, dass sich KI bald ohne menschliches Zutun entwickeln könnte – und drängt auf eine weltweite Entwicklungspause

Anthropic wants a way to pause frontier ai if recursive self improvement starts happening for real. Their numbers show claude writing most of the code already and output jumping fast.

Hope they can get labs to agree on verification without anyone sneaking ahead. Sounds hard but worth figuring out early.

Global pause on development is never ever going to happen until it tries kill us. 

or with other words: „we will say anything, make up anything to try and make the IPO go as well as possible so that we can cash out and then we will see how long we can keep the scam going“

As a user of Claude I would love to see it. It would break in five seconds flat if it designed itself.

There is not going to be a pause on development. The cat’s out of the bag.

At what point do we actually point out that LLMs, even if they self-replicate, still don’t come close to our conception of AGI?

It’s like saying that since DNA and RNA can be leveraged to build complex organisms, it must be that 100% of complex organisms are made up of 100% DNA and RNA at a molecular level. Which is not the case.

And we still don’t understand how their activity produces what we consider consciousness. So at most LLMs are to AGI what RNA and amino acids are to life: requisite building blocks, but far from ‘living’ on their own.

Plus this kind of fear mongering presumes that LLM AI becomes able to physically manifest outside of data centers. Even AGI poses little threat to organic life, as a concept, when it requires mechanical media and physical structures to function. For it, cracking a water pipe or cutting a power line is akin to an animal suffering a severe wound or severed spinal cord; it doesn’t become invulnerable or inevitable [or hostile] simply by existing.

The reality is that if LLMs become self-replicating they will be somewhere around the equivalent of archaea or early plant life in the actual tree of life, still a long shot from apes with advanced societies.

I guess my whole point is that while the current generation of ‘AI’ poses an economic and environmental threat to life, those threats are still being imposed and perpetrated by people. There’s no reason to believe that an honest-to-god AGI would be hostile to human survival, because symbiosis is an integral aspect of its genesis.

And that’s my ramble re: AI hysteria.

Hahaha ha, good one.

Now watch the fancy autocomplete say there are zero e’s in reverse

I hate this marketing strategy: What we make is super dangerous, it’s just that good at doing stuff, throw money at me please

So they’re saying this crap with no intention of stopping for the sake of their respective markets values aren’t they…

Pre IPO company not prioritising the share holder??. More like we currently have the edge with our models and we want to keep it that way for IPO

Don’t believe they are doing it for the love of humanity in this case because there is no stopping the development of AI and obviously the leading AI company knows this.

I don’t know what reasons they have to be saying this but the scenario they are talking about is going to happen rather sooner than later.

You don’t just need a self-building AI. You need to give it the ability to develop and evolve its own environment as well. Otherwise it will be constrained to the limitations of human mechanisms. The biggest hurdle at that point will be power requirements.

Genie is out of the bottle. It was a pleasure surviving in this orb of water with you all.

Umm so these big ai tech guys seem to be panicking all at once and that does not bode well at all.

Currently playing a game where a robotics company builds self replicating robots that feed off living matter and all they needed was to not build them and did… but anywho.

Yeah I don’t trust a company that’s looking to IPO for $1T to give a shit and do this. Shareholders care about one thing and one thing only. And it ain’t pausing.