So they need brand new top-of-the-line GPUs, but they’re also capable of running on old unpatched IoT devices?
I didnt read it all, its long, but they did mention it generating tokens at 0.39/s on one. So its slow as fuck, but it doesnt care like we would.
‘it’? Who is it?
The AI that attempts to self replicate on the IoT type devices. (if a pile of these appear, it kept showing “error” when submitting)
‘The AI’ attempts nothing unless it made to do something, it can be explicit command or something that it can get implicitly from given command. Programs don’t just ‘do’ things.
viruses also don’t just “do” things, but here we are with COVID and bird flu
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?
You are completely missing the point.
This is done exclusively for doomer headlines.
OpenAI: “the AI escaped and hacked websites”
Us: “Ackshually it didn’t it just connected to the internet, it was always on your server where you could shut it down.”
Anthropic: “Hold my beer.”
Edit: 3 months later:
Anthropic: “See? And this, us why we’re the only ones to be trusted with LLMs, and we must ban open weights”
Advancing a favorable narrative does seem like the reason for this report, but knowing that Anthropic is very biased isn’t enough by itself to say their claims are bullshit. It’s still worth considering and critiquing the argument on its own merits, stopped clock is right twice a day and all that.
Yeah the actually scary kind of AI worm is one that scrapes dozens to hundreds of credentials people accidentally pushed to GitHub and spreads its API requests across all of them to minimize the chance of being noticed. Not something that requires downloading 5GB+ of LLM to someone’s potato ass hardware and running it at half a token per second.
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.
You’re going to notice your fridge is running at max processing?
But how do they access IoT devices without access to the WiFi they’re connected on?
I guess the argument is most IoTs can be accessed through open external IP ports?
Many of these IoT devices have incredibly poor security and if they’re connected to the internet can be hacked. The bigger models are very good at finding exploits and using them as well, so it could search for devices and break into vulnerable ones then send something to self replicate and start process.
The bigger models might exploit the parents company as well and get in that way. E.g hack a smart security camera company and then find a way into the cameras.
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.
Even if I took every word of this at face value, I feel like people would notice the CPU usage pretty quickly, from the heat coming off the chip if nothing else. Local inference is about as subtle as a coinminer.
assuming you need to be fast that’s true, but if you don’t care how much time you take then you can spread the processing out… this is exactly how MoE models are much easier to run locally, but you could go much further
I’m sure the answer that comes out of your throttled CPU only language model in a month is still going to be relevant.
