AI Talk: Vatican, antibiotics, COVID-19 and PI

March 13, 2020 / By V. “Juggy” Jagannathan, PhD

This week’s AI Talk…

Predicting COVID-19 structure

I came across this effort on the DeepMind website. DeepMind researchers have been working on predicting protein structures—dubbed the “protein folding problem”—for a while now. They have been experimenting with AI and neural network models to predict protein structures. They have had a fair degree of success on this front. In their latest paper, they have applied their ability to predict protein structure to predicting COVID-19, the disease caused by the novel coronavirus. They have made their efforts open source so anyone can use their results to research and study this deadly disease. They do caution that their work on this front is preliminary and on-going, but given the urgent nature of the threat to humanity, perhaps half an answer is better than none.

AI and Vatican

The Roman Catholic Church convened a meeting with leaders of IBM and Microsoft. The topic of discussion? Ethics and AI. Pope Francis weighed in on this issue, recognizing the pervasive and possibly disruptive influence of AI in society. The goals of this confluence of thoughts from diverse quarters are the same as articulated by various tech companies on the topic of ethics and AI. Now it has the backing of the Vatican, which underscores a growing recognition that incorporating ethics is a fundamental building block.

ML and antibiotic resistance

I saw a reference to the role of machine learning (ML) in developing antibiotics in one of my usual sources—MIT Technology Review. This article in Quantamagazine is a fascinating commentary on how ML is being leveraged in innovative ways. Over the past few decades, it has been clear that there is a whole new class of bacterium which has developed resistance to almost all the current crop of antibiotics. The medical community has been over prescribing antibiotics for years, leading to the evolution of antibiotic resistant strains. Over 700,000 deaths are attributable to such drug resistance bacteria and that number is likely to grow exponentially without the development of new antibiotics. The new knight in shining armor here is ML. Professors Colins and Barzilay of MIT,  from Biology and Computer Science respectively, brought their respective skills to a problem—searching for molecules that have certain properties (ability to stop bacteria from multiplying) and are not toxic (i.e. not a cyanide). Leaving the parameters of search unconstrained, the search yielded unexpected molecules  that are turning out to be entirely novel solutions for the old problem. The new molecules are now close to realizing new antibiotic treatments. In honor of HAL 2000, they have named one of the new antibiotics Halicin.

PI and Bouncing Balls

This story in an oldie—but I am resurrecting it given tomorrow is PI-day! This one is about how PI shows up in mysterious places. For example, check out this YouTube video. If a big ball hits a small ball with some velocity, and there is a vertical wall which bounces the small ball back in the direction of the big ball, and if all collisions are elastic and there is no loss of energy anywhere, and the relationship between the big ball and small ball is precisely given by the following equation:

Big mall mass (M): 16*100N * small mass (m),

the number of collisions before the big ball reverses directions follows the digits of PI! Go figure. If the big ball is 16 times larger, collisions required is 3. If the big ball is 1,600 times larger, collisions required is 31. And, the relationship continues to infinity and you can unravel each additional digit of PI. The video shows the math behind why this is true as well.

I am always looking for feedback and if you would like me to cover a story, please let me know. “See something, say something”! Leave me a comment below or ask a question on my blogger profile page.

V. “Juggy” Jagannathan, PhD, is Director of Research for 3M M*Modal and is an AI Evangelist with four decades of experience in AI and Computer Science research.