Blog

Tracking A.I.’s Impact on Jobs…with the help of A.I.

14work2-master768.jpg

A panel working with the National Academies of Sciences, Engineering and Medicine has published a much-needed report on developing new tools to track two important trends:

  1. the rate at which A.I. is developing
  2. how these developments are affecting U.S. employment

As the co-chairmen* of the panel put it, we are currently “flying blind” on these trends.

Thus, without inventing some new kind of ‘radar,’ we won’t know either our location or where we’re headed, and we won’t know how to give career and training/retraining advice to vulnerable U.S. workers.

E.g., “Mr. Smith, your current job likely won’t exist in 6 years; here’s a related job that probably will, and here’s how to start training for it.”  Or, “Ms. Jones, the college major you’ve chosen most often leads to these 3 careers, which all have a >70% chance of being automated in 15 years.  Consider another major!”

As the panel notes, however, elements of these tracking tools already exist, in the form of the A.I. and Big Data infrastructures currently in place (LinkedIn, Google, etc.).   What is needed, the panel says, are public-private collaborations to combine the existing mountains of data with secure, anonymous, and unbiased ways of distributing and making sense of it.

Thus, one essential way to track and adjust to the development of A.I. is by means of A.I. — provided that oversight for the common good is also in place.  (In particular, machine learning’s focus on gleaning practical insights from petabytes of data will be key.)  If properly directed, the very technologies that threaten so many workers’ jobs may, it turns out, help put those same workers back to work.

*The panel is co-chaired by Erik Brynjolfsson of MIT, the co-author of the outstanding The Second Machine Age — a must-read on the topic of A.I. and employment.

“The Simple Economics of Machine Intelligence”

This piece in the Harvard Business Review has a four-part argument, with a cautiously optimistic conclusion:
  1. Machine intelligence is essentially about prediction
  2. As the price of such prediction drops, demand for it will go up (e.g., predicting very early-stage diseases)
  3. The ‘complement’ (in economic terms) of prediction is judgment — done by humans.
  4. Thus, demand for human judgment will also go up [e.g., decisions about medical treatment], and this is good for human employment prospects.

How should we educate, in the age of automation?

3543

In the Guardian, George Monbiot cites arguments that the dominant mode of education in the West may have been well-suited for an industrial age, but not for a post-industrial, increasingly automated one.

While reading, I thought of a few other institutions that have long sought to de-regiment, de-mechanize, and genuinely humanize education: the Montessori tradition, and St. John’s College.

Article: “In an age of robots, schools are teaching our children to be redundant.”