The case that AI skeptics are judging last spring’s models

Paul Graham shared a passage arguing that skeptics who say AI cannot conjecture, theorize or explain are describing models from last spring rather than current systems.

The passage predicts that, within months, those objections “will be describing nobody.” It also asserts that the evidence is already available, but Graham’s post does not identify particular models, benchmarks or examples. The argument is therefore a forecast about the pace of model progress, not a demonstrated comparison.