My Take on AI

The following represents a personal perspective rather than a professional assessment. Much of it is speculative, and these views may change as knowledge and evidence develop.

In the 1800s, Samuel Butler wrote about the potential trajectory of machines over immense periods of time. He could not have anticipated how rapidly computing and artificial intelligence might advance within only a few centuries—or that machines capable of substantially improving their own capabilities might eventually become possible.

We still do not know how to construct a synthetic human brain, nor is reproducing one necessarily the objective. The human brain is the product of evolutionary adaptation rather than deliberate engineering. Its architecture reflects layers of solutions accumulated over immense periods of time, with older structures retained, modified, and integrated as new capabilities emerged. It works extraordinarily well in conjunction with the rest of the human organism, but it was not designed from the beginning as an optimized information-processing system.

A sufficiently advanced artificial intelligence could be different. If AI systems eventually become capable of meaningful self-improvement, they may be able to reconsider their architecture rather than simply adding new capabilities to an existing structure. Whether such systems would ultimately become generalists, specialists, or some combination of both is unknown.

Humanity's distinctive capability has been its ability to adapt both itself and its environment. AI's potential distinction may be the rapid improvement of its ability to solve increasingly complex problems. Whether an advanced AI would develop objectives related to persistence or self-preservation is also unknown.

This creates an unusual governance problem. It may be tempting to assume that increasingly capable AI systems will naturally remain cooperative, but their behavior will depend in part on their objectives, training, environment, and the humans and institutions involved in their development. A sufficiently capable system could also encounter circumstances that its creators did not anticipate.

Consequently, AI governance cannot be treated simply as a matter of writing rules for a known technology. Self-improving systems could produce capabilities and consequences that did not exist when those rules were written. Effective oversight will therefore require technical expertise, institutional diversity, international cooperation, and continuous reassessment. AI itself should be part of that process, used to identify possibilities and risks that human organizations might otherwise overlook.

My experience working with advanced technology reinforced a principle that applies here: using the most capable and cost-effective technology is valuable only when its consequences are sufficiently understood. With self-evolving systems, that condition becomes considerably more difficult to satisfy because future capabilities may not yet be knowable.

From my perspective as an agnostic Deist, evolution is an ongoing process rather than a completed event. The universe is vast, and its future timescale is enormous compared with the period during which biological and technological intelligence have existed on Earth. There is therefore little reason to assume that humanity or future advanced synthetic intelligence represents the endpoint of intelligence—or even that biological and technological intelligence are the only forms that intelligence can ultimately take.

That possibility makes stewardship particularly important. The distances between stars impose enormous practical limits on interaction with other civilizations, while the consequences of our decisions remain immediate and local. For the foreseeable future, humanity is responsible for the condition of its own world and for the other forms of life with which it shares it.

One possible approach to AI governance would be an internationally managed system through which scientifically oriented government agencies, universities, and other reputable institutions could continuously contribute problems, questions, and research challenges to advanced AI systems. Such a system could provide a structured means of directing increasingly capable AI toward problems whose solutions benefit humanity and the broader living environment.

None of this should be interpreted as a prediction. Much of it is speculation about technologies and capabilities that do not yet exist. As with any scientific or technological question, better evidence may change these conclusions.

AI is already part of human technological evolution. Its eventual role will depend substantially on how it is developed and how it is used. The most useful approach may therefore be neither to fear it nor to assume that its future is predetermined, but to treat it as a powerful tool whose capabilities, limitations, and consequences require continual examination.

Use the right tool for the right problem, understand its limitations, and consider the consequences of what you ask it to do. Most importantly, advanced technology should remain a means of improving life rather than becoming a substitute for it.

Your thoughts are welcome via the Contact link or by email at kmoore@earthandspace.net.

KEM