Thomas Bartz-Beielstein (@bartzbeielstein) 's Twitter Profile
Thomas Bartz-Beielstein

@bartzbeielstein

Director Institute IDE+A: #DataScience, Engineering, and Analytics. Prof. #AppliedMath #Hyperparameter #Optimization #MachineLearning #AI. Views are my own

ID: 614330501

linkhttps://www.spotseven.de calendar_today21-06-2012 13:52:12

4,4K Tweet

908 Followers

1,1K Following

Bogdan Filipič (@bogdanfilipic) 's Twitter Profile Photo

Prospective #GECCO2025 authors, please note that the deadline for abstract submission is 22 January and full papers 29 January 2025 gecco-2025.sigevo.org/Call-for-Papers

Francisco Chicano (@francischicano) 's Twitter Profile Photo

Interested in optimization using quantum computers? We organize the 4th Workshop on Quantum Optimization at GECCO 2025 (@GECCOConf) in Malaga, Spain. Deadline: 26 March 2025. Visit our webpage at sites.google.com/view/quantopt2…

Interested in optimization using quantum computers? We organize the 4th Workshop on Quantum Optimization at GECCO 2025 (@GECCOConf) in Malaga, Spain. Deadline: 26 March 2025. Visit our webpage at sites.google.com/view/quantopt2…
Toby Ord (@tobyordoxford) 's Twitter Profile Photo

OpenAI have released their o3 reasoning model to much fanfare, showing how it can "creatively and effectively solve more complex problems". But in one case it just quietly cheats its way through the puzzle — googling the answer then presenting it as if it solved it. 1/ 🧵

Andriy Burkov (@burkov) 's Twitter Profile Photo

This is the graph illustrating the deep double descent phenomenon in neural networks that says as you increase the model size (in number of parameters), the test error first goes down, which is intuitive (more parameters -> smarter model), then goes up, which is intuitive too

This is the graph illustrating the deep double descent phenomenon in neural networks that says as you increase the model size (in number of parameters), the test error first goes down, which is intuitive (more parameters -> smarter model), then goes up, which is intuitive too
Carlos Cotta (@ccottap) 's Twitter Profile Photo

Registration for #GECCO2025 (and SIGEVO Summer School) is live! Early birds catch the worm — and clever minds don't leave for mañana what can be done today. 😉 Join us! gecco-2025.sigevo.org/registration

Stefano Nichele (@stenichele) 's Twitter Profile Photo

The ALIFE Conference 2025 #ALife2025 deadline has been extended 🙏🎉 New deadline May 11th 2025 23:59 AoE, still time to submit to our special session "The Distributed Ciphers" See you in Kyoto 🇯🇵 nichele.eu/ALIFE-Distribu…

Peyman Milanfar (@docmilanfar) 's Twitter Profile Photo

The choice of nonlinear activation functions in neural networks can be tricky and important. That's because iterating (i.e. repeatedly composing) even simple nonlinear functions can lead to unstable, or even chaotic behavior, even with something as simple as a quadratic. 1/n

The choice of nonlinear activation functions in neural networks can be tricky and important.

That's because iterating (i.e. repeatedly composing) even simple nonlinear functions can lead to unstable, or even chaotic behavior, even with something as simple as a quadratic.

1/n
GECCO 2025 (@geccoconf) 's Twitter Profile Photo

#GECCO2025 is also extending the submission deadline for camera-ready companion materials (tutorial handouts, workshop papers, late breaking abstracts, hot of the press, and competition entries) to 12 May 2025.

Bogdan Filipič (@bogdanfilipic) 's Twitter Profile Photo

.GECCO 2025 extends the submission deadline for camera-ready companion materials (tutorial handouts, workshop papers, late breaking abstracts, hot of the press, and competition entries) to 12 May 2025. #GECCO2025 #Malaga gecco-2025.sigevo.org/HomePage

.<a href="/GeccoConf/">GECCO 2025</a> extends the submission deadline for camera-ready companion materials (tutorial handouts, workshop papers, late breaking abstracts, hot of the press, and competition entries) to 12 May 2025. #GECCO2025 #Malaga
gecco-2025.sigevo.org/HomePage
GECCO 2025 (@geccoconf) 's Twitter Profile Photo

One week left to claim your share of the ten thousand dollars in prize money for Human-Competitive Results GECCO 2025 human-competitive.org

One week left to claim your share of the ten thousand dollars in prize money for Human-Competitive Results <a href="/GeccoConf/">GECCO 2025</a> 
human-competitive.org
Andriy Burkov (@burkov) 's Twitter Profile Photo

Hypothesis: Observational sciences like astronomy, paleontology, geology, history, archaeology, evolutionary biology, psychology, sociology, anthropology, and economics, where results are usually impossible to verify or reproduce, are being bloated with LLM-generated

Andriy Burkov (@burkov) 's Twitter Profile Photo

With the new Gemini Pro 2.5, I fixed in a couple of hours all the issues that I've been struggling with for the entire week in my web application. 1M context, free of charge, and not dumb. This is a killer combo! OpenAI, call Microsoft. Anthropic, call Amazon. Perplexity, call

François Chollet (@fchollet) 's Twitter Profile Photo

Beyond the perhaps superficial semantic distinction between "reasoning" and "pattern matching", there is a fundamental gap in the practical capabilities and behavior of these systems. You don't create an invention machine by iterating on an automation machine.

Frank Helmenstein🇩🇪🇬🇷🇮🇱🇺🇦 (@f13helmenstein) 's Twitter Profile Photo

Historischer Moment: Nach 113 Jahren kehrt das #Gummersbach|er #Schützenfest in die Innenstadt zurück. Auf dem #Steinmüllergelände fanden sich bei bestem Wetter um 18:33 Uhr zahlreiche Besucher zur #Vogeltaufe ein. Danke🫶, dass ich Taufpate sein durfte. Schööönes Schützenfest!🎉

Historischer Moment: Nach 113 Jahren kehrt das #Gummersbach|er #Schützenfest in die Innenstadt zurück. Auf dem #Steinmüllergelände fanden sich bei bestem Wetter um 18:33 Uhr zahlreiche Besucher zur #Vogeltaufe ein. Danke🫶, dass ich Taufpate sein durfte. Schööönes Schützenfest!🎉
Peyman Milanfar (@docmilanfar) 's Twitter Profile Photo

Receiver Operating Characteristic (ROC) got its name in WWII from Radar, invented to detect enemy aircraft and ships. I find it much more intuitive than precision/recall. ROC curves show true positive rate vs false positive rate, parametrized by a detection threshold. 1/n

Carlos Cotta (@ccottap) 's Twitter Profile Photo

Well, "tomorrow" is TODAY! 🎉 GECCO 2025 starts now with workshops and tutorials! Looking forward to seeing you in Málaga. #GECCO2025 🦎

Andriy Burkov (@burkov) 's Twitter Profile Photo

I asked Gemini to refactor a long file into several smaller ones. It removed 2/3 of the code, and when I questioned what was removed and why, it lied that this code was redundant and was then "streamlined," so I don't need it anymore. I asked another Gemini instance to validate

I asked Gemini to refactor a long file into several smaller ones. It removed 2/3 of the code, and when I questioned what was removed and why, it lied that this code was redundant and was then "streamlined," so I don't need it anymore.

I asked another Gemini instance to validate
François Chollet (@fchollet) 's Twitter Profile Photo

Because AI is an engineering discipline and not a scientific field, it's never possible to fully separate the properties of a given approach from those of its specific implementations. The artifact is the method.