Feedback speed

Feedback effects do not imply an explosion. Many of the definitions below, read literally, define RSI as feedback effects. Yet the authors often go on to say that RSI would have explosive implications. I think this is a mistake.

Most RSI definitions conflate feedback, autonomy, and explosion; an actual explosion requires a cardinal capability scale with a meaningful zero — a conceptual clean-up of the takeoff-speed debate.

more research inputs accelerate progress only to the extent that research can be divided into different useful tasks, carried out simultaneously, and usefully recombined.

Introduces parallelization technology as an overlooked bottleneck: millions of AI researchers may not produce an explosion unless the ability to divide and recombine their work keeps pace — a new argument for gradual takeoff.

the vast majority of frontier AI lab staff will neither sleep nor eat nor use the bathroom. They will grow smarter and more capable each month.

AI labs are already automating their own research, and this is a distinct inflection point whether or not it produces a discontinuous explosion. Part II (in the ledger) argues for audit-style third-party verification of labs.

thus far, empirical work on the software intelligence explosion has used flawed data and models.

The whole software-singularity debate rests on shaky data and should be settled empirically — the live reference for the 'measure it, don't model it' position.