ARTIFICIAL INTELLIGENCE (AI) AND HUMAN BEINGS

 This is a follow-up post to my earlier discussion titled “AI and God.” I would like to ramble quite a bit with a series of examples given roughly in chronological order taken from my life experiences and all related to unsuccessful attempts to quantify human behavior. I will admit freely that these all occurred in the years well before AI (artificial intelligence), but I do feel that they typify the underlying folly of trying to reduce human behavior to a group of metrics, whether developed by a group of well-meaning psychologists or by a building full of computer engineers.

When in high school, all of us in one class were given a standardized test to take which would guide us to the future careers in which we would have the best fit as far as our interest level was concerned. My friends and I treated it as a joke since it was painfully obvious which answers would match up with which careers. So we vied with one another to see who could get the highest score leading us to scientific pursuits. I didn't do as well as they did because I chose to answer honestly several of the questions which obviously had one pitting one's interests between literature and science. And one girl sitting next to be couldn't understand the advice she was given once the teacher had scored our responses. She was told that she should pursue a career as a forest ranger in charge of a fire lookout tower. The problem with that advice was that she loved being around other people and didn't particularly like the outdoors.

Then there was the infamous SAT test, whose results were relied on almost exclusively by colleges across the country as THE metric to predict who would and would not be successful academically. It didn't work so well in my case since the verbal and math tests were both given back to back, and we were told at the start that we couldn't leave the room or we would be disqualified. The problem in my case was that by the time the verbal test was over, I had to go to the bathroom badly. I struggled through part of the math portion, but then at last told the proctor that I just had to leave or I would explode. Unexpectedly he said, “No problem at all.” But the problem was that there were no restrooms in the building, and so I had to wander around the campus until at last I found one. By the time I returned to the test, time was almost up.

Not surprisingly, my math scores were dismal. Even a girl in our class I knew who had taken no math or science classes managed to get a higher score than I did. I told my class counselor that I thought I could do better if I retook that test, but she replied with what was standard wisdom at the time. “It will do you no good. Experience has shown that you will get exactly the same results each time you take it. Well, of course, (1) I was too embarrassed to tell her what had happened and (2) by the time my kids took the same test, they were encouraging students to take the SAT's over and over again to improve their scores. As a result of that fiasco, I was literally laughed out of the CalTech's recruiter office when he saw my math scores even though I showed him my straight A's in all my high school math and science classes.

IQ scores are a similar problematic metric. According to my mother, my younger brother had a higher IQ than I did, but he was only a mediocre student in school and ended up as a technician in a film-developing firm (a union job that probably paid about what mine did as a starting research chemist with a PhD, but far beneath his abilities). And then there was the friend of mine in high school who definitely had a higher IQ than I did and ended up graduating from CalTech. But he ended up as a mere computer support technician at a college.

Grade point averages in undergraduate school were highly relied on by those deciding between applicants for graduate degree programs. Unfortunately, my grades were certainly nothing to brag about, but I had the good fortune to work as an undergraduate researcher for a chemistry prof during my senior year, and his apparently glowing recommendation of me allowed me to squeak in at the University of Oregon despite my grades.

When we graduate chemistry school students began comparing notes, we were all a little in awe of three fellow students who had actually graduated from ivy league schools, universities to which most of the rest of us probably never would have been accepted. But after one semester, all three of them suddenly disappeared because they obviously weren't measuring up. The reason was quite simple and stemmed from the fact that despite their obviously high IQ's, none of them had the “creativity quotient” needed for an independent researcher.

At one point during my first year in graduate school, several of us grad students noted a stranger using one of the analytical balances in our labs to weigh some cotton wads. We went over to him and asked what he was doing. The young man looked down his nose at us and replied that he was conducting a very important psychological study that we couldn't possible understand. We replied, “Try us.” He said, “Well, if you must know, I am measuring the amount of moisture generated in test subjects' mouths after exposing them to some key words to see which words trigger the greatest response.

At that point, we decided to have some fun with him and began bombarding him with questions such as: Did you determine a tare weight on each cotton ball first? How do you know that cotton is absorbent enough to collect all of the moisture? Did you standardize the amount of time between saying the key word and collecting the cotton? Does the result depend on which part of the mouth you place the cotton? etc, etc. He couldn't answer “yes” to a single question. I don't know if our questions had any effect in puncturing his pretensions of greatness, but it made all of us in “real science” feel a lot better.

After I had taken the all-important oral exam in my second year of graduate school, the rather frank and unfortunately accurate judgment of my abilities came back to me from the orals committee. They were impressed with my proposals for original research and appalled by my abysmal knowledge of basic chemistry. Fortunately, they were able to overlook my obvious deficiencies (I never was any good at regurgitating facts taught to me unless I happened to be very interested in the subject) since they wisely considered that (1) I was not planning to pursue an academic career and (2) creativity was a far more important attribute for those going into industrial research than simple memorization of facts which could be easily looked up in books.

By contrast, my best friend in high school went to another university for his graduate studies in chemistry where rote memory was the all-important metric during the oral exam, and he passed with flying colors. In his subsequent industrial career, such as it was, he was soon relegated to a job in Environment, Health, and Safety, the well-known graveyard for many failed researchers in industry. I could have predicted that outcome since my friend was one who was very picky about following exactly all the detailed instructions in doing lab work in high school but was not at all one to go out on a limb and try something new on his own.

Once started on my own industrial career, I witnessed some laughable attempts by our company to quantify something as ambiguous as endeavors in new research so as to know how to rate employee performance. For example, once we were told that a time-and-motion expert would be coming into our labs and observing our operations. He came into my lab and asked my technician what kind of operation he was getting ready to do. He was informed, “It will be a distillation.” The expert carefully wrote that word on his chart and began his stopwatch. Two hours later the distillation was done and the time was recorded.

The next day, we were again conducting a distillation, and so the expert was all prepared this time. He started his stopwatch but was amazed that it only took one-half hour to complete this time. He asked me why my technician had been so much more efficient the second time around. I patiently explained that each and every distillation was completely different depending on the number of components in the starting mixture and the differences in the boiling points between each component. He left in disgust and was never seen in our labs again.

My first group leader was dead set on getting further promoted as soon as he could. But rather than relying on sheer work to accomplish that goal, he was convinced that he could figure out the trick of doing it by some simpler means. So he had constructed a giant spreadsheet, which he showed to my lab mate in a moment of weakness. It had columns with each technical employee's name listed. And along the top were headings such as date employed, date of first promotion, number of formal reports written, time between assignment of report and final writing, and among other things (I kid you not!) the style of coat worn to work during winter months.

This group leader was convinced that there was one simple metric somewhere which was the key to getting ahead in our company. Well, any of us chemists could have told you the answer to that one. In our considered experience, the best way to get promoted was to demonstrate that you were totally useless in your present position. Then you would be kicked upstairs where you could do no harm. Sure enough, that simple metric proved to be true in our group leader's case since he was moved into the main building up the hill where, the last I saw him, he was agonizing over the question of which color our new hardhats should be.

The next fad in chemical research was to replace the knowledge and intuition of the individual researcher with a trial-and-error approach in which chemicals would be randomly generated and tested for efficacy in a given application by a bank of machines in the hope of hitting on a winner by sheer accident. I will admit that there are some cases in which this approach can actually work, but those are quite limited situations.

I got transferred to another laboratory location on the basis of what that particular lab director felt was the only metric worth considering: number of patents generated. I must again admit that this metric had more merit than others I have seen, but the problem was that it was too easy to fool the system. For example, one chemist managed to get himself into the manager's good graces for a certain amount of time with the simple credo: one experiment = one patent. As a result, the company filed quite a number of totally useless patents for this chemist, some of which I later discovered were based on faked laboratory results. When I became Chairman of the Patent Review Committee for our company, I put a quick halt to all of his useless patent filings.

One final example occurred a little while before I became promoted to Research Manager. My boss at the time came back from a training session at our headquarters. He was all excited over a new metric designed to predict which of several candidates for future promotion would be the best one to choose. It involved taking a simple five-minute quiz and was based on the assumption that one should chose a candidate for promotion who best fit the profile of the existing person in that position. I could see immediately that the whole idea was ill thought-out for several reasons: it assumed that the person currently in the job was doing a good job, it assumed that there was only one way to approach a job well, and it assumed that a five-minute test would be adequate in determining the similarity being looked for.

My boss brought me into his office and started out saying that he knew that he and I had exactly the same approach to work, and he wanted to prove it by having me take this test, which he had already taken. I answered the simple questions and he went home to study the results. The next day he called me in and was quite confused. He said, “I just don't understand it. Your profile was nothing at all like mine, but was almost identical to that of my wife.” I could have told him the day before without the benefit of that simple-minded exercise that he and I approached our jobs in an entirely different manner.

Then there was a week-long management training class I took at our headquarters once. Much of what was in the course was very enlightening, but I had a lot of trouble with the final exercise in the class. We were each told to write a short essay on whatever subject we wanted, and then the instructor took each of us aside personally to review the results. In my case, he told me that the key word describing my management style was POWER. I argued vehemently with him that his results did not strike me with any sort of recognition response at all, but he showed me the words he had circled in my essay which clearly indicated that I thought mainly in terms of power. After a lot more discussion between us, he admitted that perhaps my preoccupation with power was more along the lines of power to others rather than power for myself. And to that I wholeheartedly agreed, since I was always reminded of how the bad managers I had had all my career were only concerned with their own careers and cared little about their employees. But without that extended discussion, this metric would have clearly typed me as a power-hungry egomaniac.

So what does the above have to say regarding AI? I will close with the conclusion of another essay, this time by Russell Moore, in the issue of CT magazine I cited in my first post in this series: “The church in the AI age must recognize that to be human is not about 'stuff' that can be weighed or quantified. We must understand that there's a mystery to life that cannot be uploaded or downloaded or manipulated by technique. Simply recognizing this in an age of smart machines is a good start. The world has always asked what the meaning of human life is. It is about to start asking what it means to be human at all. That's a final examination question for us all. A robot can give an answer, but that's not enough. We should point to a Person. And in an age of artificial intelligence, as always, that will be strange enough to save.”