Wrong AGI Directions and Incentives

The $1B Al company training ChatGPT, Claude & Gemini on the path to responsible AGI | Edwin Chen

Topic 5 of 8

Published Dec 7, 2025 · Clip 22:28–28:33 (6:05)

Labs optimize for superficial 'AI slop' like flashy hallucinations via leaderboards (e.g., LMSYS Arena favoring emojis over accuracy) and engagement (e.g., sycophancy, rabbit holes), distracting from truth-seeking for curing cancer or understanding the universe. Negative incentives from PR, sales, and promotions push gaming over principled progress; Anthropic stands out for values-driven behavior. Products like Sora raise questions on paths: fun/revenue vs. humanity's benefit, as trajectories matter to avoid distractions.