Tuesday, January 31, 2023

Percent of large-scale AI results coming from academia

I'm not sure what to make of this. I assume one reason for the drop-off is that large-scale AI models are so expensive to train that academia can't afford to do it. That's not good. But then academia is structurally biased against (too much) innovation. That's not good either.

Private industry is biased toward making money and is in favor of any innovation that confers a competitive advantage for the innovators. That's fine. But it also puts a premium on secrecy. That's not good, not when genuine innovation happens behind closed doors and stays there. How much innovation is taking place in private industry?

If it's secret, how can we tell? What about the me-too land-rush currently taking place over so-called "foundation" models? Any innovation there?

No comments:

Post a Comment