The Place of the Human in an AI World

The massive availability of data in recent years means that humans alone can no longer process the enormity of available information quickly and efficiently. This is why, through machine learning, we train machines to do what we can no longer do – sort through this data to identify patterns and predict outcomes.

Through Artificial Intelligence (AI), we expose the machine to stimuli from the outside world and create interactions between the human and the machine. Movie producers seem to argue that in the inevitable evolution human-machine interaction, the machines will eliminate the humans; or perhaps Hollywood producers make movies they know we all believe to be true. Serious academics and researchers – some and not all – argue that the role of the human will become obsolete in the next two generations. Stephen Hawking famously warned that AI could end all of humankind.

But can they really?  

I suggest we humans spend some serious time thinking about this future, to ensure we do not become a virus that machines try to eliminate from the earth. Machines lack one essential quality that makes it impossible for them to ever fully be able to operate without human oversight, no matter how sophisticated AI becomes: ethics. Therefore, we must not empower machines to make adjudication decisions which require the application of value judgments.

Computers do not inherently know good from bad or right from wrong. They do not know the joys of friendship and family and a good meal shared. They cannot prefer the taste of chocolate to strawberry ice cream. The closest they can come to human values would require a human to implant that preference in a machine in the form of a line of code so that they could mimic ethics.

Think of the (perhaps) billions of people interacting with a bank today; I can imagine +99% of those people engage in legitimate and legal activities. While financial crime and fraud adds up to a massive dollar amount, it comprises a small fraction of the population of bank customers.  A bank official, therefore, could benefit from an efficient machine identifying those who pose little threat, so that humans can focus on those who do.. What attributes correlate with financial crime or fraud? This machine could find patterns based upon past instances of crime and fraud, generate models, and prioritize the banking population for Anti-Money Laundering (AML) purposes, for example.

Fantastic, right! Right? We are moving into a world where machines collect every kind of data we previously never imagined, from shoe size to organic food preferences to race and sexual orientation. The benefit of a computer is that it can find non-obvious correlations between patterns of human behavior we may not have considered and the behavior we are concerned with, like defaulting on a loan, which can make our businesses more efficient. It can also be terribly terribly wrong. 

Generations before the fictional machines destroy the human race, today’s machines can cause great damage. Machines will draw conclusions that humans would know cannot be correct; in a sense Machine Learning is dumb. In addition, machines know no ethical boundaries and so will cross lines that humans would not. For example, a computer could find correlations between a person’s race and his or her likelihood to launder money. The computer would, without guilt or hesitation, make transaction recommendations based on these metrics. But these recommendations would not only be unethical; they would factually mislead the human decision maker!

Referring to examples from Tyler Vigen’s Spurious Correlations site, a computer could conclude that U.S. spending on science, space and technology causes people to commit suicide. A graph would show that the rises and falls of suicides from 1999-2009 correlated with eerie accuracy with the rise and fall in science spending. But as humans, we know this to be a coincidence. We laugh it off only because we have the ability to ignore these results. But what if the computer from the movies were to actually run the government – would investment in science decrease in order to save lives?

Movies may have us believe that AI will one day overtake the role of humans in the world. This simply is not true. Human preferences are informed by our values. We understand spurious correlations. We think about privacy and security. Computers can learn these preferences from us, but cannot independently create these values. And human values have allowed us to hold our position at the apex of earth. Unless, of course, I am wrong.