Study finds little measurable financial impact from AI spending
In his Substack newsletter, Gary Marcus cites reporting by Laura Bratton of The Information on a new study (no direct link provided) finding it remains "very difficult" to identify actual, materially measurable financial impacts from spending on AI. The piece recalls early-2023 expectations that ChatGPT would 10x every employee's productivity and boost national GDPs, but says there is still no evidence of that happening at scale, leaving a gap between documentable ROI and trillions of dollars invested. Marcus also states in the same newsletter that early generative AI darling Character.AI has filed for bankruptcy and that CoreWeave bonds have reached 15%, claims he gives no sourcing for and that should be treated as unverified.
It reopens the AI-ROI debate: if measurable financial returns stay elusive, the case for massive compute and model spending will face tougher scrutiny.