The Binary Fallacy

If you haven’t read it yet I highly recommend a new book, Artificial Unintelligence: How Computers Misunderstand the World (MIT Press), April 20, 2018. The author is Meredith Broussard, a computer scientist turned journalist who specializes in algorithmic accountability reporting.

If it sounds technical, it’s not, although there is enough technology to help you understand the issues. What there isn’t, and what I want to address here, is much about the implications of Broussard’s valuable insights for business. And they are significant.

First, however, we need to state the problem. It is, simply put, that all existing forms of digital technology are no more nor less than binary processors. That’s how digital technology works. And the implication is, as a result, that in order to automate a process it must first be converted into discrete binary steps.

Reality, however, is not always binary. A mathematical computation is. Things like inventing a new product or addressing a customer issue, on the other hand, are not. The reality of most business processes, in fact, is defined by a broad array of variables, each of which can be multi-dimensional. That is why your day is filled with problems. For every binary rule there is an exception; lots of them.

As business people we have raised technology and automation to near-religion status. It’s easy to see why. Every businessperson wants to reduce costs and more often than not that means eliminating bodies. Our infatuation, however, is misguided and it’s leading us down a path of destruction.

There are many dimensions to the problem. One, of course, is simply the way we account for profit and loss. The business P&L is not calculated in any computational sense. It is approximated using universally accepted principles of accounting. In the end few businesses really know how much money they make on which products or services. They’re only guessing.

And these rules, as they currently stand, put a high degree of emphasis on body count, in part, because the wages we pay someone are easily measured. Every ROI calculation, as a result, puts an undue emphasis on direct costs and largely ignores or estimates the indirect costs.

Here’s a simple example. I moved into a new home not long ago and, of course, wanted to get cable and high-speed internet installed. And while I knew from experience that the process would probably be unpleasant, I was not prepared for what actually happened. The cable company is not only no more responsive than it was a decade ago, before I moved to China, the process is now automated. The company is, as a result, both unresponsive and throwing money out the window.

There are two issues and both are ultimately linked to automation.

When the technician shows up at your door you can be assured that he or she will have a tablet in hand. And that tablet, you’ll soon learn, is the real boss. The technicians can do nothing unless the tablet tells them to.

But of course the tablet is not a sentient being. It is merely the user interface for a digital processor. Which is why, when I developed a small problem a few days later and approached a technician from the same company working in the neighborhood (It’s a new subdivision.), he couldn’t spare five minutes to help me. “Unless this little guy tells me to do it, I can’t. Company policy. Sorry.” So I called the 800 number and had them put it in the tablet. Another tech drove out a few days later and, as suspected, was there for five minutes.

A related but separate issue is the division of labor, which, like technology, companies have taken to an extreme, in part, because automation enables them to. (It would be impractical otherwise.) The technicians will connect your wires but they won’t bury your cable. The company hires a sub-contractor to do that. And while the techs drive a spiffy new van with the company logo on the side the cable buriers show up in a beat up old Fiesta with the bumpers held on by duct tape. And the company, undoubtedly, believes it’s actually saving money.

And because I live in a new subdivision of townhomes, every time a new unit went up another group of people came out to connect the utilities, and on three separate occasions these diverse groups, all working serially, not concurrently, accidentally cut my cable. They told me about it, nicely enough, but, of course, they couldn’t touch the severed cable. That job wasn’t in their tablet and their time is closely monitored in the interest of accountability.

That effectively started the whole process over again. It took well over two months before I had reliable internet access and by my unscientific estimate, the amount of time the various company workers spent driving to and from the job versus actually doing some value added work was about 50:1. (I won’t even touch on the impact of all that driving on the various companies’ green initiatives, which they spend a lot of money touting.) And, of course, we, the consumers, are paying for all of it in both fees and poor customer service.

There is a lot of talk of late about the disruption of robots in the future workplace. I think it’s a pipe dream even if the robotics companies do achieve scale. There are some menial, repetitive jobs that can be digitized. As a percent of the whole, however, it’s a small slice.

The jobs that contribute the highest value added, more importantly, will never be digitized. Or at least not effectively so. I do have to admit, however, that the accountants and the financial analysts have a very stilted way of looking at costs so I have no doubt that many companies will plow ahead and automate jobs that shouldn’t be.

And I don’t have time or space to even begin to get into the fact that most of this great technology doesn’t work all that well and requires boatloads of cash and technicians to maintain. What other industry on the planet has the luxury of releasing new products that it knows don’t work on the assumption that they can issue a constant stream of updates and fixes to make them function as promised?

I’m all for technology assuming it is applied properly. And it can’t be applied properly unless it is understood. The sad reality, however, is that most business executives don’t understand it. I believe the tech executives do, but I also believe that the tech industry is ruled by a culture of libertarian ideology that makes them far more tolerant of risks than the average corporate CEO can afford to be. 

That’s why they will probably put fully autonomous vehicles on the road long before they’re ready for prime time, as they have historically done with new operating systems. The tech folks don’t see risk as their issue. Their job, as they see it, is disruption. The rest is up to the user.

Language isn’t helping, as it seldom does. Artificial intelligence is an oxymoron. Machine learning is a myth. Machines can’t learn or think in any sense remotely resembling a human being. Machines are digital processors that can simulate thinking and learning through the use of recursive binary calculations called algorithms. They cannot solve complex problems; they can only predict a probable answer. 

They can do that pretty well, admittedly, with simple tasks like language translation and voice recognition, which are essentially binary. They cannot, however, innovate or be in any way flexible or accomodating. And until someone comes up with something more than the binary processors at the heart of all technology known to humankind today, they never will.

Don’t place your bets on the wrong resources. People are still the most valuable resource you have by a long shot. But if you don’t invest in them to the degree you invest in your technology they won’t be engaged. And then you’ll just be left with a bunch of machines that don’t work and nobody motivated enough to cover for them.