Institute for Advanced Study / Princeton University Joint Astrophysics Colloquium

Artificial Intelligence (AI) refers to a set of techniques that rely primarily on the data itself for the construction of a quantitative model. AI has arguably been in development for three quarters of a century, but there has been a recent resurgence in research and application. This current (third) wave of AI progress is marked by extraordinary results --- for example, in image analysis, language translation, and machine automation. Despite the aforementioned modest definition of AI, its potential to disrupt technologies, economies, and society is often presented as (nearly) unmatched in modern times, due in part to the versatility of the algorithms in modeling a wide variety of data. Similarly, there is great promise for applications across the sciences --- for example, simulations, image classification, and automated experimentation --- which are currently being investigated by researchers across the globe. Along with the significant promise of AI, comes great peril: in societal contexts, the consequences include enhanced surveillance, facial recognition, and automated weaponry. In science contexts, the issues are also significant and in many cases related --- for example, bias, lack of uncertainty quantification, and misuse. To take full advantage of the opportunities for AI to accelerate science and improve society, it's essential that we carefully guide its development.

During this presentation, we will explore modern AI techniques, like neural networks, and review how they are being developed and deployed in astronomy. Then, we’ll discuss ideas for the future usage of AI in science, including technical barriers for long-term application. Finally, we’ll discuss the roles of scientists and academic communities in the development of AI algorithms.