How many headlines have you seen blaming computers for a range of problems? Bank ATMs giving out the wrong amount of money. Utility statements with a few extra zeros on the total. Finding that you are double booked on a flight. It’s all the computer’s fault.
Except it isn’t. For anyone in who has watched the UK sketch series Little Britain, the problem is best shown through David Walliams’ recurring character of the woman tapping at a keyboard for a while before saying, “Computer says no.”
A computer is a pretty unfailingly logical piece of equipment. No deus ex machina here — everything is hard-wired or programmed in the way a computer operates. No amount of AI in commercial computer systems has allowed a computer to single-handedly decide that this time around, 1 + 1 will equal 3.
No, computers do as they are told. Yes, they will give the “wrong” answer if they are told to do things wrongly. This can be a case of errors introduced at coding time, which means that the “wrong” answer has been hard-wired in, or it can be that it is more a case of “operator error: please replace operator”.
Coding errors should be picked up during testing — if an application is so complex that exhaustive testing cannot be carried out, then is there probably something wrong with the application’s design. Of course, meaningful error messages could also help clarify, rather than further cloud, things.
Operator error is a little bit more difficult to deal with. You could try and make sure that the operator is so completely well trained and trusted that they never make mistakes. You could weed out those who cannot realize that a result they receive has to be wrong, during job interviews. Or — and here’s a different idea — you could automate things.
If the main cause of errors comes from carbon based entities (i.e. the humans), then it makes sense to transfer the power to the silicon-based entities (i.e. the computers). As computers do as they are told, then as long as the code is correct they will not make errors. Not only that, but as computers can do millions of things every second, they will do these tasks faster than a human. The end result? Happier customers, more sales, and a better bottom line.
OK… yes, I do remember the “fatal flaw” found in a certain CPU some years back, and the problems in some commercial off-the-shelf software (COSS) systems leading to erroneous calculations. Hopefully, these are now in the past — COSS worth its salt will have been checked for code quality and will have been extensively tested; on the hardware side, I haven’t seen a repetition of such design errors for some time.
In a battle between silicon and carbon as to lack of process errors, my money is on silicon winning every time.
This still leaves a fairly major place for humans, though. Ask a computer to make a decision based on a limited set of information assets, and it will still struggle. Get it to identify the odd one out from a series of pictures, and you’ll be looking at pretty expensive software that will still make errors on a regular basis. For humans, surely this is where our skills lie? And not only the skills, but the interest. I’d certainly be far more attracted to a job where I am making informed decisions based on using my brain to weigh information in a way computers can’t.
There’s hope for the humans as yet — just remove them from the low-end tasks and use them for dealing with the more complex issues.