The Future of ML Automation

Logan: We talked about how data was a bottleneck for potentially making these models much, much more capable in terms of being something closer to what we would call an artificial general intelligence. Logan: How do we get more data if we've exhausted the internet? And so one part of ML research is writing up experiments. Like people at Google, like 3% of their code, like recently was co-pilot. They use it at work at Vow.