I would say the same about CHAT and other AI functions. There is too much forging ahead, deliberately avoiding consideration of the potential consequences. The very recent public statements and letters of caution from experts in the field testify to that. And yet we see countless news articles about promoting general AI, and still others recording that these programs are already in place and working for us. It’s a matter I began to study several years ago on behalf of an ethics company where I was an associate. Experts told me that the computer programs, even algorithms, were rapidly becoming so complex, typically building on old structures, that it would not be possible to deconstruct them because the damage to them and the resultant cost of putting them back together, would be prohibitive. Also, I was told, no one person would be able to look at such a program and actually understand everything it was doing. Available would be only the choice between using it, or not.
The Alignment Problem, by Brian Christian, identifies some problems with the very material which is used to train AI, e.g., our current daily language causes some difficulties with how AI works. It is in the very language we use to communicate with other humans, e.g., the connotations we ascribe to terms of gender and sex, fields of work (some are less important than others), and to other terms which have values associated with them. Christian argues that requiring the removal of values from AI, or making preferable values, or developing AI thought without values, will be difficult until we clean up our language(s). In English a simple example would be pronouns relating to people: he, she, and they. Now that “they” is used to denote a non-binary person, for example, it may be confusing to use “they” to denote more than one person. I can only imagine what this means in languages which ascribe gender to inanimate objects and abstractions such as Greek, Hebrew, German, the Romance languages (and others with which I am not familiar).
Christian is a writer, an outsider, but his work has won awards from groups of scientists. He delves deeply into the many difficulties involved in aligning AI with human values, beginning with the problems of aligning the human values themselves, i.e., developing consensus about what they are and what they should be. His book is a very deep dive, and I am glad I read it carefully. I recommend it to anyone because the human problems aside from AI are worthy of reconsideration. But it was published in 2020, and he asserts very early in his conclusion:
The story told thus far is an encouraging one, of assured, stead scientific progress. An ecosystem of research and policy efforts to influence both the near and the long term is underway across the globe; this is still large nascent, but it is gathering steam.
But now CHAT GBT and others have arrived, and the resignations and letters of warning have arrived. The experts are reconsidering. So should we.
Another category of scientific research which, I think, needs reconsidering is the movement to bring back the dinosaurs and other pre-historic creatures (many books – Google “extinct doesn’t mean forever”). About this there are the classic ethics questions: why, who and what benefits, who or what is harmed? The discussion about these matters is immense, but the efforts continue.
But we know what we are doing, right?
We know that we are largely responsible for the rapidly more frequent and violent weather changes, and floods. Canada’s wildfires are well known even in the States, and we all look at Maui and nod with understanding when Hawaii’s acting governor said that no one expected a hurricane 500 miles away to blow such fierce winds that drive the fires so terribly quickly. We know that this is related to our cars and trucks and ships and aircraft, manufacturing, ever growing demand for energy of several kinds, our logging practices, and our economies which are based on having more of everything. Yes, there are weather changes which happen without our efforts; but we know that our efforts have exacerbated these changes. We know what we are doing.
We know that we have structured our societies so that many, many people will be unable to survive the heat, floods, and droughts in many parts of the world; that there will be long-range food and water shortages, and ravaging diseases and fungi; that much sea life is being unalterably eliminated. There will be those privileged to survive the travails of the moment; and the not privileged who won’t because we have never built our societies in ways that respect and care for all of us. We know that is what we have been doing.
We inevitably consider what-ifs, i.e., the ultimate outcomes. This can be frightening, of course. The fright is in addition to so the collapse of many of our fundamental assumptions about our safety:
our confidence in the essential rightness of democracy, and its efficacy;
our confidence that our governmental systems will hold, and have prepared for what-ifs;
our confidence that science can protect us from fungi, bacteria, and viruses;
our assumption that technology will solve the other problems;
and our confidence that we can count on everyone else to help to pull together to save us and themselves (I doubt that we will confidently expect everyone to protect themselves from the next plague).
Our religions are shakier than ever because of the damage they have done to people in the recent past: mainline North American Christan religions’ schools for indigenous children designed to assimilate them, right-wing religions which want to create theocracies, and religions in Asia persecuting each other.
Probably we do not know what we are doing. And we are surprised to have to reconsider whether we have done as well in the past as we had thought.
What comes next? It’s pretty easy to forecast the likeliest consequences of climate change and the continued wars within and between countries: severe food and water shortages, heat that kills people in short order, waters that drown them and destroy their homes and employment, and governments unable to preserve themselves or their countries. There will be oases, and probably many of us count on being in those oases. But we will (I hope) feel really terrible about those outside them and (again, I hope) will do our best to help them whether or not we actually can. Some of our efforts will in fact help, and some will not. Despair offers no benefits whatever, so we will (I hope) continue the efforts, if only to have some peace of mind, and to keep our souls from aching because of the suffering, and because we didn’t try.
Is there anything in our future on which we can rely? Yes – our values. Come hell and high water, we must live our values to the extent that they don’t harm others. Which values?
- Life – everyone’s, everything’s. None of us is wise enough to decide which life is more worthy of continuing than another.
- Compassion and mercy. To feel worthy of ourselves and of whatever divinity we worship (if any), we must try live compassionately and mercifully, and work at doing acts of kindness even when our fears threaten to drive us to anger (fear and anger are the two sides of the same coin).
- Prayer. Those of us who believe in theistic divinities may also pray that our efforts will be in the direction of divine intention and more effective than if we were just slogging it out on our own.
- Witnessing to avoid aloneness. We must bear witness to each others’ lives. Let no one be alone and unknown. Hold hands spiritually and emotionally, and sometimes physically. Touch can be a powerful aid to life.
I could elaborate on these four values, and draw examples for my readers. I refrain from doing so. Your thinking will do that. Their very general quality is likely more helpful than trying to develop specific approaches for specific cases. I expect that most of the kind of people who are inclined to read this blog would agree with these broadly named values, and will lay your own plans to see to them in your lives (as probably you are already doing). Additionally, we will benefit from talking with others about their values and ours – not to agree or disagree, but to prove to each other that we are not really alone in in these ways of looking at life.
I suspect that the ultimate results of these deliberate living our values will accomplish more good than we know to expect, and that when our times in life are ended, we
will know what it was to which we have given informed consent,
we will like ourselves, and
it will be well with our souls.
Upon reconsidering... 