Thomas Gabel

CS Prof at Frankfurt University of Appl. Sciences, focused on machine learning and AI, robotic soccer team leader.

[Thomas Gabel @nightship]

@DavidSacks @grok Name all persons in this photo from left to right!

[Jürgen Schmidhuber @SchmidhuberAI]

Against his better judgment, @CadeMetz continues to claim in the NYT [1,2] that his friend @GeoffreyHinton helped develop distillation in the early 2010s [5], even though Hinton's team merely republished — without attribution — my 1991 distillation technique [3,4]. What a shame.

[1] C. Metz, New York Times (6 July 2026). American AI companies Say Chinese Copycats are Quickly Catching Up.

[2] C. Metz, New York Times (25 Sept 2026). Is China Really Stealing A.I. From American Companies?

[3] J. Schmidhuber. Neural Sequence Chunkers. Tech Report FKI-148-91, TU Munich, April 1991. This paper introduced 3 techniques that have become essential for large language models (LLMs): (I) pre-training for deep neural nets (the P in ChatGPT); (II) positional encoding (Sec. 5.1); (III) distilling the knowledge in a neural network (Sec. 3.2.2 & 4), now widely used to train LLMs (in 1991, this was called "collapsing" or "compressing" the behavior of one neural network into another).

[4] J. Schmidhuber. Learning Complex, Extended Sequences Using the Principle of History Compression. Neural Computation, 4(2):234-242, 1992. Based on [3].

[5] O. Vinyals, J. A. Dean, G. E. Hinton (2015). Distilling the Knowledge in a Neural Network. arXiv:1503.02531. This paper failed to cite the original 1991 neural net distillation procedure [3][4].

[6] J.S. Tweet (23 July 2026): "I support open-source models distilling what commercial companies distilled for free from the entire internet. I published distillation for free in 1991 in Europe - this was copied in the US and in China." https://t.co/MBEAkF73C5

[Marc Porter Magee 🎓 @marcportermagee]

This is a better way to visualize Columbus’ voyages than the traditional way (that shows him traveling through a modern map)..

It looks a lot crazier and more daring when you see it as a trip into the unknown.

[Thomas Gabel @nightship]

Adorable!

[Bearly AI @bearlyai]

British mobile operator 02 created an AI voice bot called “Daisy” to combat phone scammers.

Daisy is an “elderly grandmother” that isn’t very good with technology. She’s a bit hard of hearing and goes on random tangets ahout knitting and food recipes.

Scammers waste time chatting with her instead of scamming real people.

***

Link: https://t.co/R2vCIvhyWT

[Thomas Gabel @nightship]

And, suddenly, 1989 feels like yesterday. @ylecun

[Yann LeCun @ylecun]

https://t.co/V6aDTm9yLP

[Thomas Gabel @nightship]

Fourth Quadrant

[Arthur Spirling @arthur_spirling]

This is basically how I think the "AI debate" is playing out. Most of my mutuals are bottom right (I am too)

[Thomas Gabel @nightship]

@zacharylipton unterirdisch

[Thomas Gabel @nightship]

Yes. Sometimes just (temporarily) changing a single (most often the last one) character in the recipient's address. Just in case of accidentally hitting Ctrl+Enter.

[Sinead Bovell @sineadbovell]

Does anyone else draft important emails without the recipients address? Read through it many times over to catch any errors. And then add the email address

[Axel @rach_prelude_c]

Bitcoin hat über 75% seines Wertes seit seinen Höchstständen verloren und verbraucht immer noch etwa 150 TWh Strom jährlich — fast ein Drittel des deutschen Stromverbrauchs. Eine gigantische Energieverschwendung für eine systematische Wertvernichtung.

[Joscha Bach @Plinz]

The most important question is not “how does the mind work?” but “how does something that builds a mind work?”

[Thomas Gabel @nightship]

Vor 25 Jahren: Kasparow unterliegt Schachcomputer – doch der KI-Ansatz verliert | heise online https://t.co/7V1unxSRoE

[Thomas Gabel @nightship]

@RiffRacerReques @hagenlaw_ @Arnd_Diringer die Lehrkraft, der Lehrkrafter

[Thomas Gabel @nightship]

Zahlen, bitte! Alle 26 Sekunden: der mysteriöse Puls der Erde | heise online
https://t.co/b7giy72TU3
Ist vermutlich ein schlafender, vergangener Gigant ;-) https://t.co/QPHo1AkmZd

[Thomas Gabel @nightship]

@Plinz I didn't read that carefully. But my first thought was: the Borg.

[Thomas Gabel @nightship]

Germany is much different: number of cases over deaths vs. number of new infections. #covid19 https://t.co/Nied0Mrjbp

[Thomas Gabel @nightship]

Frankfurt University of Applied Sciences: Entire university on home office as of now. https://t.co/VThBJ8gkkW

[Thomas Gabel @nightship]

“Flattening the Curve” has recently been around as a phrase to describe an approach for counteracting the Covid pandemia. #flatteningthecurve #COVID19 #coronavirus Here, @Plinz tries to quantify what this actually means. Worth a read: https://t.co/PeO7fC1oqC

[Thomas Gabel @nightship]

Amazing. That was 59 years ago ...

[Brian Roemmele @BrianRoemmele]

It was 1961 and in this rare film we see a discussion about how computers and computer intelligence (AI) will impact humans.

Featuring a spectacular prediction by the astonishing Mr. Claude Shannon.

In some ways we are still having this 1961 discussion.

[Thomas Gabel @nightship]

@TimKietzmann This is like mentally pressing Ctrl+Z after recognizing that you've done something really stupid.