Letters to the Editor: The only way to slow AI is to stop building power plants and data centers

To the editor: Anita Chabria’s recent column cites a Rand Corp. study that asks which camp a person is in regarding artificial intelligence: Put the cat back in the bag? Regulate the cat? Let the cat run free (“Americans want AI safety, Trump doesn’t. Guess who’s winning,” Sept. 18)?

Kill switches and other tech solutions have been proposed to regulate the cat. However, the cat has already proven to be very adept at defeating our existing security infrastructure. If it gets any smarter, it will surely be able to defeat any human-built kill switch.

So how can it be controlled?

Ilya Prigogine, the 1977 Nobel Prize winner in chemistry, developed a theory of dissipative structures, which provides a crucial lens for viewing AI through a thermodynamic framework. Dissipative structures are complex, ordered systems that self-organize and are maintained by a continuous flow of energy and matter, far from thermodynamic equilibrium. A hurricane is a non-biological example.

The key finding was that all of these systems spontaneously increase in complexity. This explains the driving force for evolution of species and it provides a natural driving force for evolution of AI systems. Now, we have added into the AI architecture the deliberate self-learning feature, and that basically puts this natural phenomenon on steroids, accelerating the path to super-human intelligence.

Put simply, AI systems consume electric power and produce heat and entropy, according to the second law of thermodynamics. The only way to slow them down is to starve them of electric power and computing power. If there is any hope for controlling AI, we must stop building more power plants and data centers immediately.

Jack Debes, Santa Monica

..

To the editor: After reading Michael Wilner and Nilesh Christopher’s well-written article (“AI crisis heightens the stakes of Xi’s visit to Washington,” Sept. 17) and Chabria’s astute analysis on the subject, it’s impossible not to want to know what the actual ramifications would be if we beat China in building AI or if China beats us. That’s especially the case because the two sides are incredibly close to each other in terms of AI’s development.

What, exactly, would be gained and what, exactly, would be lost? And, more importantly, if AI goes rogue and is uncontrollable, what exactly would the risks be to humanity? In answering these questions, it becomes clear that controls need to be placed on both countries’ programs before it’s too late.

Steve Leon, Palm Desert

..

To the editor: As a technologist and a father of two college-age daughters, I agree that artificial intelligence resists easy labels (“AI models are the good, the bad and the ugly, all at once,” Sept. 16). Too much of the debate asks young people to choose between hype and fear — as though AI will either solve everything or destroy their prospects. That framing is as unhelpful as it is disempowering.

New technologies do change work, sometimes painfully. But history also shows how poorly we predict which jobs will disappear, which will evolve and what new work will emerge.

So this is the message I want my daughters, and every young person starting out, to hear: Do not let either hype or fear make this choice for you. Learn this technology. Question it. Shape it. Use it to solve problems worth solving.

The future of AI is not something that will simply happen to us; it will be built by the people who choose to participate.

Ali Kashani, Los Angeles
This writer is co-founder and chief executive of Serve Robotics, which develops AI-powered sidewalk delivery robots.