The Institute for Progress released a collection of essays by guest authors on how AI could be used to accelerate progress in science and defend against its major risks. You might find a few familiar names among the authors, including Miles Brundage, davidad, and Sella Nevo. Here's the list of...
Applications are open for ML4Good Colombia April 2025 In partnership with AI Safety Colombia, ML4Good is running an intensive 10-day bootcamp focusing on upskilling in deep learning, exploring governance, and delving into conceptual topics for individuals who are motivated to work on addressing the risks posed by advanced AI systems....
This project was developed as part of the Swiss Existential Risk Initiative Research Fellowship under the mentorship of Alexandre Variengien. The use of the first person throughout the document refers to Alejandro Acelas. I designed three challenges meant as a training ground for improving your skills in analyzing transformer’s internals....
The Institute for Progress released a collection of essays by guest authors on how AI could be used to accelerate progress in science and defend against its major risks. You might find a few familiar names among the authors, including Miles Brundage, davidad, and Sella Nevo.
Here's the list of essays:
Applications are open for ML4Good Colombia April 2025
In partnership with AI Safety Colombia, ML4Good is running an intensive 10-day bootcamp focusing on upskilling in deep learning, exploring governance, and delving into conceptual topics for individuals who are motivated to work on addressing the risks posed by advanced AI systems.
This bootcamp will fast-track your deep learning skills, inform you about the current landscape of AI Safety agendas, connect you with like-minded individuals for potential friendship and collaboration, and accelerate you towards taking concrete next steps towards working impactfully in this field.
The bootcamp is aimed at people in Latin America with some coding experience who hope to improve their technical and conceptual understanding in order to work on AI... (read 285 more words →)
This project was developed as part of the Swiss Existential Risk Initiative Research Fellowship under the mentorship of Alexandre Variengien. The use of the first person throughout the document refers to Alejandro Acelas.
I designed three challenges meant as a training ground for improving your skills in analyzing transformer’s internals. Each challenge centers on a transformer model trained on an algorithmic task and, depending on the challenge, solutions are either a labeling function or a file with model weights. You can submit your solution to be automatically scored on how accurately it resembles a held-out test set at the corresponding CodaBench competition:
In brief, the challenges are: