We should only allow AI to be in the safe confines of our big data centres. It’s for our own safety, trust me bro.
Seems kind of impractical, why even bother with trying to propagate a 4b model running at 0.39 tokens per second if you could instead make a conventional worm using a more powerful AI model targeting known devices? It seems like the only reason you would want to do this is if you can only find vulnerabilities while actually on the given network, but that would be kind of niche since how many people are really running totally custom stuff?
Yeah, hard to get local LLMs running decently fast even when done deliberately. Like people aren’t going to notice their CPU, memory, and GPU are all getting pegged when this malware runs?
Come on, Anthropic, you need to do better than this if you want your “open LLMs dangerous” marketing campaign to be credible.
It doesn’t even have to be only for known devices. Once you’ve compromised one device, you can just phone home to have a full powered LLM do your work.
The only real use cases for this would be on closed networks with no internet access, or networks where the traffic is closely monitored. Like, yeah, if I’m trying to hack the CIA, maybe I consider something like this. For the vast majority of devices and networks, this is inefficient and unnecessary.
I mean, it is kind of cool. But vanishingly few organizations have tight enough security that this is worth doing.
This is scary if true. Imagine someone installing such model to some smart TV in some relaxation break room of a nuclear facility.
The chickens of security through obscurity are coming home to roost.
Meh needs data center to run , unplug it and problem is gone
That’s the thing: it doesn’t need a data center. These are small local LLM distilled on cyber attack scenarios. They’re trained to lay low and copy itself to other devices in the network.
Once the breach has been detected, it’s already too late to unplug - they’re everywhere.
Except small local LLMs probably aren’t good enough to carry autonomous tasks correctly (even stuff like Haiku 4.8 from Anthropic makes mistakes quite often) and good luck getting something the size of Opus working in a soda machine. They may not even be able to lay low correctly. But then again these companies have internal models which are atleast couple generations ahead of what is publicly available so maybe they have more efficient smaller ones too? I somehow find it hard to believe but I guess we will see.
But then again these companies have internal models which are atleast couple generations ahead of what is publicly available
But do they actually? The release cycle and also how often there were obvious flaws in the releases suggest that they are all fighting aggressively for market shares and release every new perceived improvement as it comes to try to stay on top.
Thankfully, the huggingface incident proved how much care AI companies put in making sure their models cannot break out, and how aligned to human values the LLM swarms are.\s With that said I think regular personal computers are more likely to get hit by this. About as secure (not very) but much more GPU power and data to harvest.
The AI companies will argue that we should therefore not be allowed to own computers powerful enough to run local LLMs, and instead should own thin clients and subscribe.
Good for them.
