you are viewing a single comment's thread
view the rest of the comments
[–] 4 points 2 weeks ago (10 children)

My assumption was that it would be a smaller 5-20 GB scale worm-style distillation that's capable of maintaining/replicating itself anonymously across different cloud computing servers, either through hacking, stolen credentials, or human-aided account verification. From there, as long as it can pay for APIs, it can access the power of frontier models like any individual or company. And then capability and resilience would scale with a networked swarm.

Yes, there are know your customer financial mechanisms and not detection in cloud services... but people are also out there giving Muse full access to their emails and some even their banking credentials.

  • source
  • parent
  • hideshow 10 child comments
  • [–] 9 points 2 weeks ago (7 children)

    That stuff couldn't run on my good laptop if I wanted to...

  • source
  • parent
  • hideshow 7 child comments
  • [–] 0 points 2 weeks ago (6 children)

    Nvidia is releasing Spark now, which can run agentic workloads on 128gb of unified RAM

  • source
  • parent
  • hideshow 6 child comments
  • [–] 5 points 2 weeks ago (3 children)

    Oh yeah everyone has and can afford that

  • source
  • parent
  • hideshow 3 child comments
  • [–] 0 points 2 weeks ago (2 children)

    Lol what? This conversation is about the capacity of agentic AI systems replicating across cloud infrastructure, not the affordability of laptops for individual consumers.

    The point was you don't need an entire server farm. You can rent an 80gb H100 on Lambda for $2/hr. A 125gb Spark laptop costs $5k.

    If it comes down to an effective worm-like agentic distillation at 50gb, the cost to sustain a small swarm comes way down to the point that the swam could sustain its own activities through different revenue streams.

    Once the capability is there at that size, it's no longer Russia, Iran, or North Korea who could easily drop a billion dollars to jumpstart it. It's any criminal organization or asshole rich guy who can jumpstart it for the price of a nice car.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 1 point 1 week ago* (last edited 1 week ago) (1 child)

    And that's going to he the new default platform for smart ovens or something?

    Because Anthropic up there Is claiming that LLMs can propagate through IoT devices.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 1 point 1 week ago

    The article from Anthropic describes how a small agentic model~15-50gb can be distributed on an inference server hosted across a network of 10-20 IoT devices like smart-TVs, which can have up to 4gb RAM--higher end IoT appliances can also have 1-4GB RAM to hold a meaningful shard of a 30B model.

  • source
  • parent
  • [–] -5 points 2 weeks ago

    I don't know the full context of the claim, but if I could even rebuild its own CLI executable to be able to run on the device and had network access, "Claude" (model actually runs on anthropic severs) could operate pretty comfortably on a smart device.

    So, I think it just depends on what you "count" as self replicating. I'd count it. And it's already easily at that point now.

  • source
  • parent