Google, Microsoft, and other computer program providers have announced the dawn of the age of artificial intelligence. They claim that machine and programming capabilities will increase exponentially over the next few years. Computers will not merely imitate human beings in terms of initiatives and innovations but will outperform them in every conceivable field from new theories in astrophysics to original recipes for salad dressing. Programs are already available that can write original stories, compose elegant or awful music, forge college term papers, translate materials from hundreds of obscure languages to thousands of unfamiliar dialects, and imitate the actions, idiosyncrasies, and speech patterns of celebrities. Systems that drive cars, direct weapons, and facilitate moon shots are old news. Our newly achieved artificial intelligence can design cars, select targets, and redirect launches to Mars and beyond without humans intervening in the decisions.
Whether this emerging capability is good or bad is irrelevant. It is here, and it will evolve. Science fiction writers envisioned this achievement decades ago, and the scenarios they believed might develop with this breakthrough were routinely bleak. The problem that inevitably arises is that the machines take information to ‘logical’ conclusions and what is ‘logical’ is dependent upon the precedents and priorities built into the machines. There is the extreme view, as proposed in the Terminator movies, that the logical conclusion reached after reviewing all human history and capabilities is that people should be eliminated from the planet. The less extreme scenarios leave humans subservient to or entirely dependent upon the machines.
Developing machines that can think for themselves does not guarantee humanity a care-free future. Instilling information in these machines inevitably provides them with very human traits, including bias and the inclination to judge. Within a few years we shall have surgical robots that can perform appendectomies, vasectomies, tubal ligations, and other commonly performed abdominal procedures without the need for human oversight. The advantages are obvious: the machine does not fatigue or have a bad day. It can keep track of hundreds of important measurements during the procedure and can react in a microsecond to any worrisome deviations from normal. It can tolerate extreme measures needed to sterilize surgical implements, well beyond the tolerance of human hands. It can move more quickly than a person and complete the procedure flawlessly in just a few minutes. The obvious question is, “Are there any disadvantages to handing over these tasks to machines?” Based upon my medical experience, the answer is “Yes.”
A gynecologist colleague investigated a young African-American woman who had tried and failed to get pregnant for several years. She had moved to Boston from a state in the deep South and had an unremarkable medical history, aside from having undergone an appendectomy when she was a child. His investigations revealed that she had had an appendectomy, but she also had bilateral tubal ligations. The surgeon treating her when she was a child had sterilized her.
It is more likely than not that the physician intent upon this women’s never having children was a zealous racist. It is also more likely than not that this was not the only black child that he operated upon that he sterilized. Handing over such procedures to machines does not eliminate the discrimination that motivated this evil man to commit this crime against the most vulnerable of our citizens. A programmer with the same racist mentality could insert instructions into the machine’s orders that added unnecessary and unwanted procedures to the surgeries performed on specific categories of people. All this could be done in seconds and go undetected, as was the case for this woman, for years or decades.
The problem with artificial intelligence, at least in terms of medical assessments and care, is that programming starts with people. These machines will be able to program themselves, but inevitably human foibles and prejudices will be included in the information and perspectives with which they start. Any prohibition against causing injury to a person or a specific group of people can easily be subverted.
Even contributions from great minds will not necessarily lead to great decision-making or even procedures for great decision-making. We look at many judgments by people or committees or even nations and wonder how they went so far astray from common sense or even self-interest to adopt actions that were so obviously idiotic or nefarious. The brilliant Supreme Court Justice Oliver Wendell Holmes Jr. will forever be cited for stupidity in his supporting the Virginia law that legalized the sterilization of the “feeble-minded.” He stood by the sterilization of Carrie Buck, a woman who allegedly became pregnant after being raped, based upon arguments that she, her mother, and the daughter who was the product of that rape were “feeble-minded.” Oliver Wendell Holmes Jr. declared, “Three generations of imbeciles are enough.” The mother was deemed feeble-minded because she had a child out of wedlock and was probably severely hearing-impaired. Carrie’s daughter did well in school and made the Honor Roll.
If a person with the brilliance and foresight of Oliver Wendell Holmes could be so wrong in assessing circumstances that would affect tens of thousands of people who were sterilized by state order after the Buck decision, what is the probability that designers of our fledgling artificial intelligence will get it started in a direction that will inevitably benefit us? Remember, at the advent of the Industrial Revolution there were predictions that the standard of living for all would improve immensely. We ended up with a wealthy elite, blighted cities, and factories packed with children working for subsistence wages. At the dawn of the Atomic Age, there were predictions that energy production would be so inexpensive there would be no need to charge consumers for it. We ended up with expensive electricity, Three-mile Island, and radioactive waste that will litter our countryside for generations. I certainly hope we have a better result with artificial intelligence, yet another gift from the titans of technology.
Dr. Lechtenberg is an Easton resident who graduated from Tufts University and Tufts Medical School in Massachusetts and subsequently trained at The Mount Sinai Hospital and Columbia-Presbyterian Medical Center in Manhattan. He worked as a neurologist at several New York Hospitals, including Kings County and The Long Island College Hospital, while maintaining a private practice, teaching at SUNY Downstate Medical School, and publishing 15 books on a variety of medical topics. He worked in drug development in the USA, as well as in England, Germany, and France.
