I listened to an NPR show about AI,(http://www.npr.org/2017/04/21/524702525/jeremy-howard-will-artificial-intelligence-be-the-last-human-invention) and there is a lot to be concerned with. There is a concept called “deep learning” where computers can learn from their experiences, and it is getting literally better day by day. This isn’t the next industrial revolution, this isn’t the steam engine replacing animals or the assembly line replacing craft made objects or even things like robotics on an assembly line, these systems are for real and they have the potential to replace a lot of jobs. Siri isn’t what I am talking about, or driverless cars, they are relatively simple, I am talking systems that literally can learn and grow over time, and they can do so faster than human beings can. They had a humerous thing where the IBM Watson team let Watson surf the internet by itself, and in a day or two its responses to questions were peppered with obscenities, but it does tell a bit of the story.
But a lot of it raises real questions, something few people are addressing. Many of the geeks working with this stuff waxed on about how it would free people from the need to work, how it was going to create a new world where for example places in the third world where medical care was rare would have the ability to get medical care they need thanks to AI and robotics, others made the claim about ‘new jobs’, but the problem is what if in the next 20-30 years most jobs can be done via AI? And we aren’t now talking about those who aren’t educated, this could take a lot of jobs that today people may assume will never go away.
A company in my building is a well known financial firm, and they laid off something like 15% of their analysts and fund managers, because they are using more and more automated systems to make those decisions. Deep learning systems can use the same formulas that the finance types use in evaluating takeover targets or stock analysts analyzing a company’s performance, they can run Dupont Ratios and run the same analytics models that humans do, and learn from actual company performance to what they expected and tailor things. (It will be interesting to see when AI starts hitting the beancounters like this, if suddenly automation instead of being “we understand people have lost jobs over this, but this is the nature of change”, as they have said to blue collar jobs being lost, to suddenly being a bunch of whining about “but I have an MBA, how can I be redundant?”.
In that same show the last guy who spoke, who is an expert on the technology, basically said that the claims of economists and the geeks promoting it that this will create new jobs was bluntly, dubious, that this was not the same thing as the information age or earlier revolutions, this one was computers able to replace people in large scale, and that even things like programmers and ‘data scientists’ could become obsolete, that a machine learning system can learn to do those things, and for example if you want to design a new product, you can give the system a set of parameters and it could code and build it for you, and the ‘analyst’ specing it could be a machine as well.
How far this will go and how fast no one really can predict, and no, it won’t happen tomorrow, but it likely will happen, and eventually people will wake up and realize that no matter how educated they are, how skilled, they aren’t protected. A robotic AI system could replace a surgeon or a doctor, and there are already systems build that for a limited number of illnesses and issues, can diagnose a patient faster and more precise than human doctors can.
Some will say this will lead to a Star Trek world, but I suspect it is going to be a bumpy ride and it will take severe dislocations before anyone decides they have to do something. In a Star Trek world things people need to live would be free or nearly free, but what happens before then? Right now the dominent economic model has been to create orders of increases in efficiency, which in turn generally boils down to cutting down the cost of labor, so with AI you have the potential to replace a wide swath of costly labor with machines, which will make efficiency experts and finance types happy, but how in the world do you maintain a demand economic model when you are basically removing those who create the demand? How do you convince those whose basic job is to create more and more with less and less labor, to increase profits, that they may, just may, have to start thinking questions like who will buy their products? And more importantly, if the system more and more benefits less and less people, how do you create an environment that replaces the feeling of worth people get from working with other things? I am cynical enough to believe that those that benefit from reduced labor the most will hang on to the current model as long as they can, and it won’t be a smooth transition. We already see this with low or semi skilled jobs that mostly have been replaced by automation or will be, which is easy to dismiss as being ‘there are always losers, like buggy whip makers’, but what about when it hits the jobs most of us do?