[polygamous_bat]

> Why not just emailed the original authors for the raw data?

Industry research labs, especially Google deepmind, are notoriously closed up about their “proprietary” data. I’ve hit this wall multiple times in my own work in AI.

[sp332]

https://twitter.com/borgeaud_s/status/1780988694163321250 says they're going to open the data from the paper. Not sure why they didn't do it before, but good news.