Import AI 475: Swarm scaling; Google DeepMind watermarks biology; and the AI science economy

Key points

• Toby Ord has a nice, short post about how to think about swarms in terms of AI capability development. “A good way to see AI swarms is as a new form of inference-scaling,” he says.1
• Stepping on Toes parameter: Swarm scaling doesn’t scale perfectly - rather, it seems like as you increase the number of agents in a swarm you get a diminishing returns property which matches with things economists have observed about coordinating large groups of people, where you pay some kind of tax as you scale the number of people involved.1
• However, swarm scaling is still powerful enough that it suggests swarms could increase the chance of an RSI-driven intelligence explosion rather than reduce the chance. “I’d hoped that the value of λ for AI agents would be lower, making an intelligence explosion less likely, but that appears to not be the case”.1
• Why this matters - new factors in scaling: So far, AI has mostly scaled capabilities through picking the right combination of compute and data for a trained model, then figuring out how to spend inference budget on thinking via ever-longer chains of thought and tool calls, etc. Agents introduce a new parameter in scaling AI capabilities.1

Sources (1)

• [1]Import AI 475: Swarm scaling; Google DeepMind watermarks biology; and the AI science economy
  Import AI (Jack Clark) · Oct 5, 12:32 PM

  Import AI runs on arXiv, cappuccinos, and feedback from readers.

  Toby Ord has a nice, short post about how to think about swarms in terms of AI capability development. “A good way to see AI swarms is as a new form of inference-scaling,” he says.

Extractive summary: sentences quoted from the sources.