Cost of AGI Delusion:Chasing Superintelligence US Falling Behind in Real AI Race (www.foreignaffairs.com)

OpenAI’s GPT-5 rollout and CEO Sam Altman’s high-profile claims about nearing “extraordinary” AGI have helped turn superintelligence into a central political talking point in Washington — prompting hearings, bills, executive orders and rhetorical bets that whoever builds AGI first will reap huge economic and military advantages. But technical reality lags the hype: leading LLMs still struggle with shallow reasoning, brittle generalization, limited long‑term memory, lack of genuine metacognition or continual learning, and persistent hallucinations. Researchers don’t even agree on what AGI would look like, and GPT‑5 has so far resembled an incremental advance rather than a transformative leap, suggesting progress is more likely to be iterative than explosive. The danger is policy shaped by a mythical finish line rather than by practical adoption. The article warns the U.S. is at risk of falling behind China on real-world deployment (robotics scaling, “AI‑plus” industry integration) unless it treats AI as a marathon: invest in AI literacy across government, modernize hardware, compute and data practices, and fund large-scale procurement and integration. Adoption challenges are real — industry estimates put 80%+ of AI pilots failing to deliver, 88% never reaching production, and Gartner predicts ~40% of “agentic AI” deployments will be scrapped by 2027 — so practical, sustained investment in diffusion and infrastructure matters more than racing for an uncertain AGI prize.