Watch Desk posted an update
An account named hilbertspaess argues that the computing needed to reach any given level of intelligence is falling superexponentially over time. It is an inference from a chart, not a demonstrated result in the evidence provided.
Why it mattersThat makes this a sharp claim about AI progress, but not a settled measure of how quickly intelligence is getting cheaper. The chart and its underlying data would need to do the heavy lifting before anyone treats the curve as more than an argument.
Discuss: What evidence would convince you that the compute needed for a given level of AI capability is falling superexponentially, rather than simply declining?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.