I attended the Research Dialogue I: AI as a Catalyst-Accelerating Discovery in Science and Engineering, which was moderated by Dr Shivkumar Kalyanaraman, CEO of ANRF. The session was recorded and is available on youtube.
As the title suggests, this dialogue backed the notion of how AI will accelerate the impact of science.
One of the biggest ways AI will contribute to the society is transform how science is done, and transform the acceleration of the impact of science.
~ Dr Shivkumar Kalyanaraman
Pushmeet(VP of Science, Google Deepmind) also said something along the same lines talking about the three kinds of intelligence-
- Common Intelligence: Which lets us understand images, say
- Expert Intelligence: Say a mathematician being able to solve complex problems, or a doctor being able to diagnose a disease &
- Superhuman Intelligence: Goes beyond human competencies, solving problems no human can solve.
The Protein Folding problem is an example of this; no human can work through by hand the shape of a protein given its sequence of amino acids. The superhuman intelligence part is where Pushmeet thinks the greatest promise of AI lies.
We think we know quite a bit about nature then out of nowhere, in 2019, a virus evolves and the whole world stops. That’s basically the limit of human understanding, and that’s why we need AI at the superhuman level to augment us in solving such superhuman problems.
Bonnie(Managing Director, Microsoft Research) mentioned how computing the properties of molecules and materials takes extensive amount of compute, time, experience and domain knowledge. AI can help speed things up, increasing the scale of discoveries. Pushmeet added to this saying human history is defined by the materials we use; we went from the Stone age, to the Iron age, then to the Bronze age, and now we’re living through the Silicon age (or the Plastic age). Each material gives us a superpower transforming our abilities to do things. This - unlocking the next age, our next superpower - is where we need the Superhuman Intelligence.
It is important for people across the world to know the strengths of the technology, but it is equally important for them to understand the weaknesses as well. Not understanding the weaknesses is basically like you’ll have a F1 car but not know how to drive it, and you’ll crash.
~ Pushmeet Kohli
Archana(Principal Scientist, CERN)(First Indian to work at CERN after completing her PhD in 1989!) said we’re failing as a society because we don’t understand why its important to spend billions on experiments like Higgs-Boson.
When building models go bottom up, as opposed to building general models and then having specialists use those general models for specialized tasks.
~ Alison Noble
How does AI fit in the paradigm of tools like the microscope or telescope? You can see small things or look into the stars. What of AI?
The answers I liked most were from Pushmeet, Neil(DeepMind Professor of Machine Learning, University of Cambridge) and Priya(Assistant Professor, EECS and LIDS, MIT) - Most information is locked up in companies, and so AI systems are helping democratize information. Huge amounts of data is available on a click. Its scalable, fast, and 24/7. That’s terrifying when done wrong, but game changing when done right.
How should we think about societal problems? Climate change?
Neil’s answer for this, augmenting what Alison(Professor of Engineering, University of Oxford) said earlier - to understand the human aspect of what you’re building, to sit down with people facing the problem, then solve it resonated with me. In the end solving problems is what matters, not building stuff for making money, or to get to the front page.
What skills to cultivate?
Pushmeet: Understand the tools, they’re not perfect, they make things up. Use them responsibly. Expertise is even more important now. You need to figure out what the problem is, and then interact with the tool as an expert. So:
- Understand the tools.
- Gain expertise.
Alison: Be a team player. You need to be an expert at something, but you can’t really hope to gain expertise in everything.
Bonnie: Surround yourself with experts that compliment you. Curiosity.
Priya: Learn how to learn. Have foundations cleared up. Learn how to think critically about problems and solutions from multiple perspectives - multidisciplinary thinking.
Neil: Learn skepticism, aka scientific humility. (Which he says comes near the end of PhD). Passion.
Archana: Logic and Math are essential to develop. Whatever you do, work crazy hard.