Is PBT Really Cool, but in Oxford?

Grouping Oxford's PBT relies on having a vast number of complete resources. And so in Oxford, we surprisingly have quite limited for peer resources. So instead of like randomly serving high parameters by scaling it by some arbitrary value, instead you could sample anywhere from full distribution and maybe that would mean you could shrink the population size still get the same performance gains. That one was exciting because first year, I ran the experiment to my laptop,. But we showed it to all these toy environments like my beloved bipedal walker.