How much faith does AI deserve?
In the first section of this article, I discussed a recently developed software program that plays chess. In less than 24 hours, given only the rules of the game, it taught itself to play with such expertise that it defeated all comers, even chess grand masters. We learned it defied conventional rules, made counterintuitive moves, yet essentially kicked ass with all challengers. (Forgive the slang. The chess term is checkmated.)
Returning to my earlier and now antiquated software background, I trust you can appreciate that this event staggered me. I hope it will stagger you as well.
We humans have created a world of AI software, a world that, with each day, is entering into the mainstream of our lives. AI has fantastic capabilities, especially when used with the massive databases open for access to the public.
AI is also called deep learning. As one expert puts it (Foreign Affairs, July/August 2019, 192-198): “Deep learning has its own dynamics, it does its own repair and its own organization, and it gives you the right results most of the time. But when it doesn’t, you don’t have a clue about what went wrong and what should be fixed.”
Do you find this assertion bothersome? I do.
Let’s consider the modern-day issue of humans attacking other humans with nuclear weapons. Left to its own, who can guess what AI might be capable of a few years from now? Will we have the wisdom to build AI platforms dealing with weapons of mass destruction to control their self-taught logic which, like the chess-playing program, will likely exhibit opaque operations?
Will we have the capability to control an AI machine’s manipulation of genes as DNA alteration technology progresses by leaps and bounds?
What happens if, as stated above, “…it gives you the right results most of the time.” But on those occasions when it does not, then what? We cannot recall the launched missile. We may not be able to undo the genetic alteration. We cannot even determine the code and logic that went into the missile launch or a gene cut-and-insert operation.
If these questions seem far-fetched, keep in mind that AI is in its embryonic stage. Like genetic engineering, it is at the tip of an iceberg.
However it evolves, artificial intelligence will have a transformative effect on the human race. Whether we guide AI, or have AI guide us is, at this time, still up to us.
Do you think I am over-reaching? I ask you to re-read the material about the self-taught chess program cited in the first article in this series.
Also, consider this observation from the experts writing in Foreign Affairs: “…no one, not even the people who design these tools, really understands how they work.” And: “…AI technologies, at their most advanced levels, do not merely assist human knowledge; they surpass it.”
Sleep tight, Chicken Little. Hello, Hal.
- • •
Uyless Black, residing in Coeur d’Alene with wife Holly, has written “The Nearly Perfect Storm: An American Financial and Social Disaster,” available on Amazon.com.