Graphs and Machine Learning - What's Changed?

Moen: What's changed over the last couple of years? My understanding is that hardware has also made a lot of things that were not possible before. And in particular, it seems we have essentially super-computers that make it plausible to train these enormous language models. Moen: The cost for basically flop is almost a half-an-year-over-year. But trends and videos accelerators are nearly increasing in false by quite a bit year over year.