Gary Marcus calls Huang's AGI declaration "a takeover of a scientific question by corporate fiat"
Marcus on AI (Gary Marcus)·medium signal
In a September 6 post, Marcus writes that Huang "gave no evidence and no definitions" and measures Astra against his own 10-item AGI benchmark, granting it autoformalization and possibly reliable coding while doubting it hits "any of the other eight." He points to user reports that Astra performs comparably to Claude Fable 5.1 rather than representing a leap, and closes with "When real AGI arrives, we won't need to squint our eyes." This is the sharpest named-source rebuttal of the AGI-declared framing so far.