Lightnews — Scholar-powered news

Jörg Dietrich

@joergdietrich.com

Cosmologist turned data scientist and engineering manager. Working on ML systems to support human experts in their tasks.

Posts Replies Media Videos

Reposted by Jörg Dietrich

Brent Toderian

@brenttoderian.bsky.social

“Now Scientific American has updated and re-released what it calls the “classic graphic” that shows that a human on a bicycle—able to coast, or freewheel, without pedaling—remains the world’s most energy effecient traveler.” @carltonreid.com on the re-release of the iconic graph in @forbes.com.

October 17, 2025 at 12:52 PM

Reposted by Jörg Dietrich

Thomas Dietterich

@tdietterich.bsky.social

Beautifully succinct statement of the core problem with the blind application of statistical learning in AI!!

A model trained with a frequentist objective using i.i.d. observational train and validation data is almost surely converging to the non-causal model of the underlying data generating process.

That is because using short cuts/spurious correlations leads to a lower loss than not, i.e. you are actually specifically incentivizing your model not to be causal.

"Emergence" believers' core fallacy is the assumption the causal model would have the lowest error, and therefore better optimization would eventually converge to the causal model.

The opposite is true - the best scoring model (per frequentist objective) is actually the one that is best at exploiting spurious correlations.

July 20, 2025 at 8:46 PM

Reposted by Jörg Dietrich

Paul Noth

@paulnoth.bsky.social

Mine in today’s @newyorker.com

March 12, 2025 at 2:46 PM

Reposted by Jörg Dietrich

Ulrike Franke

@rikefranke.bsky.social

My latest piece for @internationalepolitik.de on the German debate on security and defence. I feel quite strongly about this one, so I hope you’ll read it.

internationalepolitik.de/de/von-krieg...