The $100 Billion Bet on Better AI Data

Scaling up AI models is no longer enough to improve their performance. We are seeing models get better at math and coding while they struggle to learn other basic skills. This shows that we are hitting a wall with current training methods.

Andrew Ho left OpenAI to start a new company focused on fixing this problem. He believes that future progress depends on how we collect and curate data. Instead of just feeding computers more random internet text, companies need to source high quality specialized information.

Ho predicts that the industry will spend over $100 billion on these targeted data sets. Getting the right data is becoming more important than simply having the most computing power. This shift will likely define the next generation of artificial intelligence.