Jörg Dietrich's Avatar

Jörg Dietrich

@joergdietrich.com.bsky.social

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

65 Followers  |  211 Following  |  15 Posts  |  Joined: 10.10.2023  |  2.4373

Latest posts by joergdietrich.com on Bluesky

Na ja. Das ELT baut man ja nur, weil das Overwhelmingly Large Telescope (OWL) nicht sicher technisch umsetzbar war.

03.11.2025 19:44 — 👍 2    🔁 1    💬 1    📌 0
Heat map of travel efficiency, updated

Heat map of travel efficiency, updated

“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.

17.10.2025 12:51 — 👍 663    🔁 204    💬 28    📌 33

Nur wenn es zum Nachtisch Kinderschokolade gibt.

09.10.2025 19:07 — 👍 1    🔁 0    💬 1    📌 0

The Time Traveler's Guide to Medieval England hat zumindest ein Kapitel die Rolle der Frau und ist allgemein ein empfehlenswertes Buch.

30.08.2025 12:58 — 👍 2    🔁 0    💬 2    📌 0

Ja, beim Kochen, wie es laut Folge möglich sein sollte.

23.07.2025 09:17 — 👍 2    🔁 0    💬 1    📌 0
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.

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.

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

20.07.2025 20:46 — 👍 48    🔁 7    💬 5    📌 1

Angst vor Abmahnung wegen § 202 Abs. 2 BGB

24.06.2025 19:35 — 👍 8    🔁 0    💬 1    📌 0

And we only need the whole freaking Earth to hold us down.

15.06.2025 17:23 — 👍 1    🔁 0    💬 0    📌 0

Isn't that simply a case of "the plural of anecdote is not data"?

04.05.2025 08:30 — 👍 0    🔁 0    💬 0    📌 0

HAM-BUR-GER
HAM - Hamburg
BUR - Hollywood Burbank
GER - Rafael Cabrera Mustelier

30.04.2025 13:42 — 👍 1    🔁 0    💬 0    📌 0

"People without a living space". Somebody's using the full Nazi playbook.

31.03.2025 13:39 — 👍 0    🔁 0    💬 0    📌 0
Post image

Mine in today’s @newyorker.com

12.03.2025 14:46 — 👍 1197    🔁 395    💬 23    📌 32
Preview
Von Kriegen, Ängsten und gefährlichem Halbwissen Militärische Fragen als gestrig abzutun, während die eigene Sicherheit von Verbündeten garantiert wird – das war lange deutscher Usus. Doch die Zeiten sind vorbei, in denen wir uns einen solchen Luxus...

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...

07.02.2025 10:17 — 👍 161    🔁 46    💬 13    📌 13

Dick Laurent is dead.

16.01.2025 21:40 — 👍 0    🔁 0    💬 0    📌 0
Picture of the moon barely visible behind clouds.

Picture of the moon barely visible behind clouds.

Impressive occultation of Saturn and the moon by clouds

04.01.2025 17:39 — 👍 1    🔁 0    💬 0    📌 0

Das sieht aus wie ein Grünfink Weibchen. Fliegen auch gerne in Schwärmen.

14.12.2024 19:01 — 👍 1    🔁 0    💬 1    📌 0

Hi, I'm Jörg. I'm Head of AI for R&D at an automotive supplier. I'm particularly interested in human-in-the-loop AI systems and organizational change.

16.11.2024 19:00 — 👍 5    🔁 0    💬 0    📌 0

Same here. Categorization in an automotive R&D context.

29.10.2024 05:57 — 👍 2    🔁 0    💬 0    📌 0
Preview
Do the Health Benefits of Cycling Outweigh the Risks? Although from a societal point of view a modal shift from car to bicycle may have beneficial health effects due to decreased air pollution emissions, decreased greenhouse gas emissions, and increased ...

Despite the increased accident risk, cycling is expected to be a net health benefit: pmc.ncbi.nlm.nih.gov/articles/PMC...).

28.10.2024 14:25 — 👍 1    🔁 0    💬 1    📌 0

@joergdietrich.com is following 20 prominent accounts