Causal reasoning and compositional reasoning are two core aspirations in generative AI. Measuring the extent of these behaviors requires principled evaluation methods. We explore a unified perspective that considers both behaviors simultaneously, termed compositional causal reasoning (CCR): the ability to infer how causal measures compose and, equivalently, how causal quantities propagate through graphs. We instantiate a framework for the systematic evaluation of CCR for the average treatment effect and the probability of necessity and sufficiency. As proof of concept, we demonstrate the design of CCR tasks for language models in the LLama, Phi, and GPT families. On a math word problem, our framework revealed a range of taxonomically distinct error patterns. Additionally, CCR errors increased with the complexity of causal paths for all models except o1.

In this post, we summarise the first two invited talks from the International Conference on Machine Learning (ICML). These presentations covered the fascinating topics of drug discovery, and the cryosphere. In Daphne's talk, she outlined some of the work she has been doing on transforming drug discovery using digital biology. To introduce the topic, Daphne described drug discovery as an interesting space that one can view as glass half-full or glass half-empty. The half-full version is demonstrated by the amazing advances in new medicines, such as vaccines, cell therapies, genetically targeted therapies, and cancer immunotherapies.

Artificial intelligence has been making waves in recent years, enabling us to solve problems faster than traditional computing could ever allow. Recently, for example, Google's artificial intelligence subsidiary DeepMind developed AlphaFold2, a program that solved the protein-folding problem. This is a problem that has had baffled scientists for 50 years. Advances in AI have allowed us to make progress in all kinds of disciplines – and these are not limited to applications on this planet. From designing missions to clearing Earth's orbit of junk, here are a few ways artificial intelligence can help us venture further into space. Do you remember Tars and Case, the assistant robots from the film Interstellar?

Artificial intelligence has been making waves in recent years, enabling us to solve problems faster than traditional computing could ever allow. Recently, for example, Google's artificial intelligence subsidiary DeepMind developed AlphaFold2, a program which solved the protein-folding problem. This is a problem which has had baffled scientists for 50 years. Advances in AI have allowed us to make progress in all kinds of disciplines – and these are not limited to applications on this planet. From designing missions to clearing Earth's orbit of junk, here are a few ways artificial intelligence can help us venture further in space.

Today is International Women's Day, and we're celebrating by highlighting some of the leading ladies in AI, machine learning and deep learning who have spoken at RE•WORK Summits and dinners over the past 12 months. Whilst the technology industry is seeing more and more women in top roles, the gender imbalance is still clear. At RE•WORK, we're passionate about encouraging women and girls into STEM and are proud to host our series of dinners and our Podcast celebrating women in AI. Earlier this year we hosted the AI Assistants Summit in San Francisco, where 50% of our speakers were women. This was a fantastic showcase of diversity, and we are striving to have more and more female experts presenting at our Summits.