AI for Science Workshop

Artificial Intelligence offers the promise of revolutionizing the way scientific discoveries are done, and tremendously accelerate their pace. In the past few years, grand challenges which were thought to be unsolvable -- or decades away from being solved -- have seen major advances using AI ranging from protein structure prediction to automating drug discovery. However, major challenges still remain in this nascent field of AI for Science, and the goal of this workshop is to address and discuss those challenges:

Challenge 1: Towards novel methods for AI for Science. The progress of AI methods has mainly been driven by tasks in computer vision and natural language processing, and the same techniques are currently used in tackling scientific problems despite being a very different domain. For example, a specificity of a large number of scientific problems is the strong presence of symmetries and invariances - how can we incorporate such inductive biases in our AI models?

Challenge 2: Tackling the right set of scientific problems. While open problems in Science abound, not all are equally likely of achieving major progress using AI. Which one should we target in priority? What makes a problem particularly suited for being tackled using AI techniques?

Challenge 3: Enabling scientific discoveries with AI. Achieving high-impact scientific discoveries using AI is usually the result of a difficult and lengthy process, involving scientists, but equally important, a robust compute infrastructure, and skilled software engineers. What are the main ingredients for doing scientific discoveries using AI?

Challenge 4: Journey from scientific discovery to a practical application. We will hear from experts on their experience of transitioning from a scientific discovery, which was made possible using AI, to a practical application with a real business value.

Confirmed speakers

Organizing team

Sponsors