“Wiener on Innovation” – Panelist — Frank Levy

Download Now HTML Transcript

“Wiener on Innovation” – Panelist — Frank Levy

Frank Levy: This is a pretty eclectic panel. I was given instructions to talk for about 10 minutes on Wieners anticipation of the current robot revolution and nits impact on any quality, so I’ll try and do that briefly.

I think a good place to start is a paper that Wiener wrote 1949, so that’s 65 years ago. It was written before The human use of human beings, but had some of the same ideas, and it was written at the request of the New York Times who wanted some sort of survey about what computers would mean for society. He wrote a draft and sent it into the New York times and the Timers went back at him and said, “Well, this is fine but there’s some modifications we’d like,” and so on, so forth. He wrote a second draft and sent it back to the Times.

The Times came back and said, “Actually there’s sets of ideas in the first draft and sets of ideas in the second draft that we’d like to combine. We’d like you to resend the first draft because we’ve lost it.” By this time Wiener was in Mexico and of course he didn’t have the cloud and he didn’t have email, and he said, “Actually, it would be quite a task for me to get my office to dig out the first draft, wherever it is, so why don’t we just drop the whole thing,” which they did.

There were something called the third draft of this essay which was put into archives at MIT and not rediscovered until 2012, when it resurfaced because of a researcher had been looking for something else. To understand what Wiener did and did not anticipate, it’s useful to take the economist shorthand for looking at skill levels in the labor force. Where an economist typically think it’s three skill levels, which you might call skill… Obviously you can talk about the subjectivity which you can associate with, say college graduates. Middle skill which has to do with blue collar workers and say community college graduates. Then unskilled workers, think of fast food workers or janitors and so on. There’s a lot of pejorative in the use of the terms, but that’s the way it appears in the economic literature.

In this essays first paragraph Wiener says, “The tendency of these new machines is to replace human judgement at all levels except the fairly high one,” but if you read the essay more carefully, by the end of it it’s clear that Wiener’s really focused on the replacement of factory work per se. He says, “These new machines have a great capacity for upsetting the present base of industry and of reducing the economic value of routine factory employment to a point where the routine worker is not worth hiring at any price. If we combined our machine potentials of a factory with the valuation of human beings on which our present factory system is based, we are in for an industrial revolution of unmitigated cruelty.”

What Wiener was doing was building on the ability of a computer to do highly structured work, which is still to a certain extent true today. He says, “It is however quite clear that apart from taping…” by which he meant writing out the software on a piece of tape, “…which is a job for an intelligent man rather than for a daft man. The apparatus on which we shall depend in the future, [machinage 00:03:26] is largely repetitive and will be capable of being manufactured by the methods of mass production.”

Wieners discussion of replacing daft men in economic terms, that I just mentioned, is really the example of replacing middle skilled workers and creating inequality by having displaced middle skilled workers compete with lower skilled workers for a lower skilled jobs, the jobs that you can’t replace in fact in today’s economy. In practice, that description is a pretty good description of what in fact has happened over the last 20, 25 years or so.

There are three other aspects of what’s happened over the last 25 years that Wiener did not anticipate, at least in this essay. Let me just briefly summarize them before concluding. The first thing is that because Wiener was writing right after world war 2, he talked really exclusively about manufacturing and didn’t mention services until he got into The human use of human beings, a couple of years later.
Even when he was writing however, the service sector employed over half of all US workers, that is in 1950. A lot of service work involve the same clerical work and so on and so forth, involve the same kind of highly structured routine information processing that made it possible to use computer substitution there as well. The whole set of office productivity work, Microsoft Office, so and and so forth, they’re good examples of how that’s really now to the user, as opposed to secretaries and so on and so forth.

To take a more recent example which sort of blends newer and older technology, one example in healthcare is the job of coder, which is somebody who reads a doctors report and then determines what CPT codes and billing codes are sent to the insurance company for reimbursement. Now, say in the case of radiologists where that report is dictated, there’s now software that can do natural language processing of the dictated report of the event of what the radiology image says and it can extract what the CPT codes are, so to process that immediately.

You have two effects of that. One is to eliminate the coders job per se, and the other is to start doing monitoring of the radiologist on a morning and afternoon basis about, “Well what did you turn out today?” How much revenue did you generate on a day-by-day basis. You see that sort of stuff spreading throughout the whole services. Service sector is just as vulnerable as manufacturing and that’s the first point.

The second point, that I think was sort of a reflection of the zeitgeist of the early 1950’s, is that quote I read before, “The tendencies of these new machines is to replace human judgement on all levels except the fairly high one.” In the economic terms I mentioned before, it sounds as if Wiener is saying, “Well, these machines are going to replace both mid skilled labor and low skill labor.” That really was the zeitgeist of the time.

There’s a famous story at MIT. Gerry Sussman is currently a professor of computer science at MIT. When he was an MIT undergraduate, Marvin Minsky who was one of the giants of artificial intelligence, took Gerry and a couple of other undergraduates and said, “Look, here’s some research money, I want you to solve the problem of machine vision. This is June, you should be able to do it by the end of the summer.” That’s kind of what people thought at that time. For those of you who have looked at this stuff, machine vision is still extremely crude and there are a whole bunch of things, it turns out, that are very easy for humans and are very hard for computers. That wasn’t really seen in the years immediately following world war 2.

The last of the three things was that the focus of Wiener is the interest that the way that we all begin to focus on this stuff, is to think about the direct substitution for people. That is to say that you think about what a job is and so to say, “How can the computer do that job?” What that misses is that in a lot of situations, jobs are affected not because of direct substitution, but because computers change the whole industry.

An example of that would be the job or journalist. There are a few programs now that can put together an article based on facts, sport articles and so on. That really has no impact on journalist jobs so far. What has had an impact is that classified advertising has moved in large bulk from newspapers to things like craigslist, so the whole economics of the newspaper industry has been overturned. There’s been a lot of shrinkage among newspapers and journalist jobs have gone with them. That’s a kind of external way of thinking about how computer revolution changes occupations, including high skill occupations and not just direct replacement of the job.

Having said that, at the same time Wiener got a lot of things in this early unpublished essay right. He briefly touched on the possibility of machine learning and what that might mean. Then there is this quote toward the end of the essay. He says, “Finally, the machines will do what we ask them to do and not what we ought to ask them to do. In the discussion of the relationship between man and powerful agencies controlled by man, the [gnomic 00:09:00]wisdom of folktales has a value far beyond the books of of our sociologists. There is general agreement among the sages of the people of the past ages, that if we are granted power commensurate with out will, we are more likely to use it wrongly than to use it rightly. “Here in a penciled edition, “More likely to use it stupidly than intelligently.”

These are not words that you hear today in most computer departments and there are certainly not words that you hear in Silicone Valley. If they came from an average person, the person would be dismissed as a Luddite, but Wiener is raising exactly the right point. While robots and computerization have not developed as fast as Wieners supposed, they continue to chip away at humans comparative advantage. Faced with that trend, we have to begin asking much more carefully what kind of society we want. Thank you.