Which of these two government policies is most appropriate for the next 50 years?
It’s a question we’ll need to wrestle with soon. It appears that we’re at the beginning of another great wave of job destruction. The last wave, starting roughly in 1980, eliminated or outsourced blue collar and clerical jobs. We used to have secretaries; now we have word processing software. We used to have factory workers; now we have robots.
The next wave will eliminate white collar jobs. This will happen in two ways:
Type 1 – through advanced communications and software support, a small number of “augmented knowledge workers” can do the work of thousands of traditional knowledge workers.
Type 2 – machines and systems will become smart enough to replace many knowledge workers.
I’ll illustrate with two examples from my life.
Type 1: The MOOCs. Massive Online Open Courses find very talented professors and augment them. With video, web, and online testing support, these professors can literally reach thousands of students. They give great lectures. (You should watch them). Why do we need other professors to cover the same material? A few professors can replace thousands. By the way, this will also accelerate the dominance of English.
Type 2: Automated Essay Grading. I’m rather proud of my ability to read essays and make useful comments that help students think more clearly and communicate more effectively. So what? Within the next few years, we’ll see software systems that can do almost as good a job as I can. OK … maybe they could do it even better since they never get tired. I’ve always thought that this would be a difficult task to automate because it’s “fuzzy”. But computers are mastering fuzzy logic even as we speak.
Much of what we call “knowledge work” is actually easier to automate than essay grading. Any process based on rules is fairly easy to computerize. Deciding which stocks to buy or sell is a good example. It’s just a set of rules. So today, “quants” and high-speed computers dominate much of our stock trading.
Diagnosing an illness may be another good example. Today, as many as 15% of diagnoses made by humans are wrong. But diagnosis is just a rules-based process. Surely, a computer can do better.
Within the next three decades, we may well reach a point where nobody needs to work. So what will we do? Good question. Perhaps we should ask a computer.