Updated forecast from Casey Briggs for a state-by-state guesstimate:
I recently had the opportunity to actually play around with the OpenAI Beta API a little. I'd previously seen some examples, along with commentary from others who have explored the model with more depth. Personal interaction, though, caused me to reflect on how much I would personally be willing to pay access to the model --- and how much value I think could be extracted from people like me in the short term. Both factors caused me to update pretty significantly.
@ThirdEyeOpen An update down. I was previously at 50%. (Sorry for the lack of clarity!)
— edited by galen
@chrisjbillington I agree with the entire picture you've painted here, and I think I was failing to really think properly about the 'latency vs throughput' distinction; thanks heaps for clarifying your position. With recent developments on the AstraZeneca concerns, I've become more pessimistic, but only slightly.
This recent study looks at the effect of whip use on race 'integrity', horse speed, and horse movement by comparing races in Great Britain where whips were carried but not used ('hand and heels' races) to races where whip use was permitted. The authors find no statistically significant difference on any front.
— edited by galen
@alexrjl "Metaculus: Come for the COVID forecasting; stay for the useful analogies in a discussion of competing desiderata for a scoring rule."
Given that the metaculus community is at 10% for deadly conflict in the South China Sea in 2020, I couldn’t bring myself to go any lower.
@admins this resolves as 2.257k.
This is a minor correction, but --- based on my reading of both the graph provided on the top500 website, and the historical data provided in the spreadsheet --- it seems as though the claim that "the United States peaked in November of 2008 at 67.02%" is untrue: an earlier US peak of 68.31% occurred in 2005.10.01.
This comment was originally posted on June 2021