sphelps.net
Scientist and CTO applying behavioural economics to AI Safety.
https://sphelps.substack.com/
https://sphelps.net/
https://www.lesswrong.com/users/phelps-sg
68 posts
61 followers
119 following
Prolific Poster
comment in response to
post
Churchill visiting the WH in 1942, during wartime, not wearing a suit
comment in response to
post
OK, not a one-to-one comparison, as Americans have no fascination with the quarter-zips. But people who insist that Zelenskyy should have worn a suit should answer a very simple question. Would Trump and MAGA have a different position on Ukraine if Zelenskyy wore a suit this entire time?
comment in response to
post
As a general matter, I also dislike the idea of respectability in dress because the concept is often unevenly applied. It bothered me to see Vance smirking during this moment, as he supposedly champions the forgotten working poor who are most often shamed because of these concepts.
comment in response to
post
en-gb.facebook.com/help/2245628...
comment in response to
post
chng.it/Y5698JRfZ4
comment in response to
post
chng.it/Y5698JRfZ4
comment in response to
post
Comments on the paper
Embers of Autoregression: Understanding Large Language Models Through the Problem They are Trained to Solve.
open.substack.com/pub/sphelps/...
comment in response to
post
> wouldn’t simply egg them on.
With a pre-registered study this is taken out of their hands. The scientist uploads their data and the LLM writes up the results. The scientists are then given the opportunity correct objective inaccuracies. Any hype introduced at this stage is recorded (changelog).