AI for Natural sciences
The experimental sciences are generating year after year an increasing number of data and information about the universe and our world. For instance, the amount of information produced by many of today’s physics and astronomy experiments is overwhelming the capacity of analysis of a scientist or even of a team of scientists.
Many research group are looking at artificial intelligence for processing and analyzing their data. With minimal human intervention, AI algorithms such as artificial neural networks can process large amount of data, discovering anomalies and matching patterns that were previously unspotted.
In particular, machine learning has become a key tool for researchers in very different domains. For example, machine learning techniques are used for: using genomic data to predict protein structures, understanding the effects of climate change on cities and regions, finding patterns in astronomical data, etc. Some scientists and philosophers argues that AI and the advent of generative adversarial networks could also generate a new scientific method, no longer based on Galileo’s scientific method, but based on generative approaches. Basically, “We don’t know anything; we don’t want to assume anything. We want the data itself to tell us what might be going on.” [cite Kevin Schawinski].
Examples of discussion topics:
- How can machine learning help integrate observations of the same system taken at different scales?
- Is there a way to incorporate existing theory/ knowledge into a machine learning algorithm, to constrain the outcomes to scientifically plausible solutions?
- How can AI be used to actually discover and create new scientific knowledge and understanding, and not just the classification and detection of statistical patterns?
Tentative workshop contents:
- Introduction and welcome
- Keynote presentation
- Presentation from participants
- Open space workshop