How do machine learning researchers actually feel about GPT-3?
How do traditional machine learning researchers feel about GPT-3?
When asked, many point to GPT-3 not being a breakthrough because it’s simply a scale up of known existing transformer models. They’ll acknowledge GPT-3 has crossed some chasm when it comes to zero or few shot performance, but not really go beyond that.
However, as a GPT-3 developer (ie. not