Ex-OpenAI researcher bets $100 billion will flow into training data because scaling alone won't cut it

Former OpenAI researcher Andrew Ho and Cambridge scholar Adam Hunt contend that scaling laws are reaching diminishing returns, as current models increasingly specialize in coding and math while stagnating in other domains. To address these performance bottlenecks, the researchers argue that the industry must shift its focus toward higher-quality training data. They estimate that as companies struggle to improve model versatility through architecture alone, investment in specialized data curation will reach $100 billion.