How misguided privacy rules could wreck the AI revolution: My review of ‘The Master Algorithm’ by Pedro Domingos

Confession: I’m not a regular viewer of Xi Jinping’s annual New Year’s address from his office. China watchers are, however. But they don’t just pay attention to what China’s president says. They also scour the bookshelf behind him. It includes both Communist writings — “The Communist Manifesto” and “Das Kapital” — and selections from the Western canon such as “The Odyssey” and “The Divine Comedy.”

Image: China / CCTV

This year there were a couple of relevant additions considering China’s push to become the global leader in artificial intelligence: “The Master Algorithm” by Pedro Domingos and “Augmented” by Brett King. Now I just did a podcast with Domingos, a computer scientist at the University of Washington. It should be up in a week or so. But I can’t credit Xi for turning me on to “The Master Algorithm,” which came out in 2016. I missed it until Amazon recommended it to me after I bought the book “Why Information Grows” by Cesar Hidalgo (which I have also reviewed and wrote a column in The Week about). And for that, I can credit Amazon’s machine learning algorithm, the version of AI that is the subject of the excellent TMA.

The book begins with a helpful explanation:

Machine learning is something new under the sun: a technology that builds itself . . . and at its core, machine learning is about prediction; predicting what we want, the results of our actions, how to achieve our goals, how the world will change.

So, yeah, Domingos is an optimist about what machine learning can accomplish — beyond it’s already considerable achievements — if scientists could create the master algorithm that combines the various versions of machine learning, a sort of Grand Unified Algorithm of Learning:

If it exists, the Master Algorithm can derive all knowledge in the world — past, present, future — from data. Inventing it would be one of the greatest advances in the history of science. It would speed up the progress of knowledge across the board and change the world in ways we can barely begin to imagine.

It’s an important aspect of the book: Domingos tells a story of a better world, one that is a welcome counter to the common dystopian vision of technological unemployment and AI running amok. You might, for instance, have a digital avatar to help negotiate with the world on your behalf. So a superpowered version of that famous Google Assistant demo making a salon reservation, one that could do your taxes, dispute improper credit card charges, and find you a new job. And, of course, your digital half would interact with other models: “If you’re looking for a job and company X is looking to hire, its model will interview your model.”

But for all that and more to happen, it will require data. Machine learning, as Domingos writes, is “an engine that turns data into knowledge.” And on that subject, Domingos makes a point often neglected during the current freak-out over Facebook’s data leak. Locking down data in the name of privacy has a trade-off. Data must be shared if science is to progress. Domingos thinks “laws that forbid using data for any other purpose than the original intended one are extremely myopic.” Such rules hinder the scientific discovery process, which may not have occurred to the writers of the EU’s new General Data Protection Rule, which kicks in tomorrow.

Of course, companies have an obligation to make sure people understand what is happening with their data. Domingos is prescient here:

Today most people are unaware of both how much data about them is being gathered and what the potential costs and benefits are. Companies seem content to continue doing it under the radar, terrified of a blowup. But sooner or later a blowup will happen, and in the ensuing fracas, draconian laws will be passed that in the end will serve no one. Better to foster awareness now and let everyone make their individual choices about what to share, what not, and how and where.