Unreal Directions

There is an article in Slate Magazine about the possibility of having AI on boards of directors for corporations. There is, apparently, one such director, in Hong Kong, that is supposed to evaluate life sciences companies, to see which are likely to be a good investment. Such directors are not currently legal in the United States, but there is nothing that I have seen that would prevent members of a board from using AI as advisors. The author of the article makes the point that a mere advisor, not having a vote, could be ignored by other directors, whereas a voting member would have to be addressed directly.

Reading the whole article, though, makes me think the technology is really not ready for the big room. The AI in use currently often can't explain how decisions are reached. Sometimes, the programmers who create the systems end up programming in very human biases. Now, having recognized that weakness, programmers are trying to create expert systems that can identify the biases, but how do we know the bias hunters won't also have biases?

I am almost interested enough in AI to look into it more, and try to learn enough to understand how it works. Almost.

One problem with an AI director is that it would be hard to sue, I expect, and if it couldn't explain how it reached its conclusions, how would plaintiffs show that it had violated fiduciary responsibilities? An outcome other than what shareholders desire is not sufficient cause for finding the artificial director liable. Plaintiffs would have to show negligence or conflict of interest, or something, which could be very hard if no one, not even the AI director, knows how it works.

There are other developments in tech. There are warnings in news stories about use of robots in health care. Robots and other AI systems only know what was programmed into them and what data they acquire. It seems that isn't always sufficient to get good results. I have wondered about that. Doctors and nurses are taught to make evaluations of patients. I would not expect a robot to have the same capabilities in noticing something not quite right in the biological or emotional realm. That shows my human bias, but I know that visual systems of robots are not the same as human vision. Can a robot tell if a spot is the wrong color? That's a judgment based on contrast, and not absolute shade. There are probably other subtle signs of difference that human health care professionals aren't consciously aware of that affect judgment and decisions. It's possible that an expert system would also make use of such things, but only within the body of data to which it has been exposed.

I am still skeptical of AI. I know that algorithms can be very useful in certain applications, but I don't expect them to be taking over this month. Or next. Predictions can be difficult, especially about the future (A quote sometimes attributed to Niels Bohr.)