In 2025, I ended up reading 43 books. Just like every year , it is time to summarize and reflect on all the books. My six favorite (☆) books of the year were: Lord of the Flies (2.) Life 3.0 (8.) Guns, Germs, and Steel (13.) The Methods of Rationality (14.) Normal People (23.) The Emperor of All Maladies (36.) Compared to the previous years, I ended up reading significantly more fiction this year,…
Google DeepMind just released the new Gemma 3n family of models. They are open-weight models, support multimodality natively, and come in sizes of 4B, 12B and 27B parameters. The developer guide has some more details on that. Something cool to look for in the announcement video (at 0:47) is the computer control demo that I both worked on and recorded: The demo also shows some fun pictures that I…
2024 is coming to an end, so it is again time for me to reflect on what I have read this year. I ended up reading 39 books this year, of which most were non-fiction, and many were health-related. Just like every year , I will summarize all the books I have read. My five favorite (☆) books of the year were: The Idea Factory (9.) I’m Off Then (11.) Situational Awareness: The Decade Ahead (14.) Good…
This post describes my work at Google: https://blog.research.google/2024/03/social-learning-collaborative-learning.html
A fun thing I recently learned about Large Language Models ( LLMs ) is that they understand base64, a simple encoding of text. Here’s a demonstration: the base64 encoding of What is 2 + 3? is V2hhdCBpcyAyICsgMz8= . Passing that to an LLM, say ChatGPT or Gemini, answers the question perfectly: Both ChatGPT and Gemini understand base64 questions And that isn’t just a one-off. It works remarkably…
As the year is coming to an end, it is time for me to reflect back on what I read this year. Just like last year , I ended up reading about a book a week. That’s a pretty good pace for me, and one that I plan on keeping next year as well. My five favorite (☆) books of the year were: Nonviolent Communication (3.) Sophie’s World (11.) The Cuckoo’s Egg (14.) The Undoing Project (30.) Outlive (45.)…
This post describes my work at Google: https://ai.googleblog.com/2023/03/distributed-differential-privacy-for.html
A habit that I have had for a few years is that I try to focus each year on some particular goal. In the past years these have been things such as learning to cook well or deeply understanding how to invest. For 2022, I decided to focus the year on reading. Concretely, I decided to read a book a week. Following the tradition of the past years , this blog post summarizes all books I have read this…
Much has been written about how federated learning (FL) can enable training much better models and help build more intelligent systems. This essay instead discusses a more personal aspect: Why working on FL as an individual is interesting, meaningful and fun. You can either take this as an advertisement for why one should work one FL, or as a personal reflection on why I am still passionate about…
For the past four years , I have been doing yearly reading summaries now. Initially, I started because I really enjoyed reading such lists from other people. Now, I mainly write them because it is a great way to reflect back on the year. This year is no different in that regard. Software Engineering at Google In 2016, Google published a book about Site Reliability Engineering, the SRE book . Last…