Sign In

Communications of the ACM

ACM TechNews

An Algorithm Might Save Your Life: How the Amazon and Netflix Method Might Someday Cure Cancer

View as: Print Mobile App Share:
Algorithms recommend movies, books, dates  even job candidates. In the future, they might cure disease.

University of Washington professor Pedro Domingos says machine-learning algorithms might one day have a transformative effect on humanity.

Credit: Shutterstock/Salon

Machine-learning algorithms underlying Amazon and Netflix recommendation engines might one day have a transformative effect on humanity, writes University of Washington professor Pedro Domingos in his book, "The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World."

Domingos says keeping machine learning opaque raises the specter of abuse and error, which makes it essential to understand machine learning and its capabilities so it is more controllable.

Not all learning algorithms follow the same operational pattern, and there are consequences inherent in those functional differences. For example, Netflix's recommender uses knowledge culled from subscriber tastes to suggest obscure films and TV shows instead of more popular choices, as customers' subscriptions are not enough to pay for choosing blockbusters. By contrast, Amazon's algorithm simplifies logistics and taps buyer preferences to direct customers to more familiar products.

Machine-learning advocates can be categorized into those that subscribe to different visions of a general-purpose master learning algorithm for discovering knowledge in any subject, although the real thrust of their efforts is a universal algorithm that can extract all knowledge in the world from data. This Master Algorithm could theoretically be the key to building domestic robots or curing cancer. In the latter case the algorithm would be able to sequence a tumor's genome, determine which drugs will work against it without harming the patient, and possibly design a new drug tailored for that patient.

View Full Article


Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account