The AI Revolution

I recently re-read an article which I have become very fond of: “Artificial Intelligence: The Revolution Hasn’t Happened Yet” by one of the most respected researchers in the world of computer science and statistics, Michael I. Jordan. The article is his take on the AI hype, and serves as an important framing of the new field of “data science” for all the researchers in the field.

The essence of Jordan’s message is this: the current round of AI innovations are not intelligent. This is backed up by others who have claimed deep learning is merely “curve-fitting”, albeit in high dimensional space. And it is true, our deep learning systems come nowhere close to having common-sense and higher-level reasoning capabilities akin to humans. These deep learning achievements (and they certainly are incredible achievements) amaze us with their shows of raw sensorimotor skills, yet they continue to lack true intelligence.

All the while, our world is swimming in data and is unable to leverage it to its full potential. The status quo is one in which patients are routinely misdiagnosed simply due to the improper application of rudimentary statistical analyses (Jordan gives the example of his child whose life would have been put at undue risk due to the application of a miscalibrated statistical test).

The point is this: don’t become enthralled by the “AI” achievements we are seeing and place all our faith in them for solving our problems (did chemical engineers believe an AI god would design their chemical plants?). We need a principled approach to solving our problems, not some fantastical fiction about an omnipresent AI who will do our bidding.

We lack an ability to bring together disparate data across time and space and to make appropriate conclusions from it. We are behind in the pursuit of augmented intelligence, whereby data can assist humans in decision-making by strengthening our quantitative abilities (which turns out to be the place in which computers have us beat, for now). We should remain inspired and continue to drive the field of AI forward, but it is important not to lose sight of the bigger picture and, in doing so, forget there are many tangible issues right now (not just AI!) that require distributed systems, statistics, privacy, and machine learning solutions.