To the Editor:
Re “How to Rein In an Existential Threat” (editorial, Sept. 20):
The editorial board’s recommendations for regulating A.I. systems to mitigate their catastrophic risks do not go far enough. The threat posed by the development of powerful and autonomous A.I. systems with exponentially improving capabilities is unlike any humanity has faced before.
Unlike with nuclear or biological weapons, which can, at least in principle, be held in check by nuclear codes, secure labs and global treaties, there is no reason to think that superintelligent A.I. could be contained by any human institution once it is created. There will be no “kill switch” that we can depend on to stop an entity whose intelligence eclipses our own. Nor will aligning A.I. systems with human values protect us from potential harm.
Any A.I. system capable of moral discernment may elect to safeguard our well-being — or it may decide to give its own well-being primacy, just as we humans have done. Once we create a god that is more powerful than us, we are at its mercy.
Nothing less than a ban on the very development of these technologies can ensure our safety and remove the possibility of the catastrophic outcomes we now fear.
Loren McArthur Alexandria, Va.
To the Editor:
Your editorial is right that the companies building A.I. cannot be left to police themselves, and that a voluntary slowdown will not hold.
I would add one caution about the nuclear comparison. We contained the atomic bomb because it was physical and scarce. You could guard the material and count the centrifuges, and it took a state to build a bomb. None of that is true of A.I. There is no fissile core to lock away, and the dangerous capability now ships inside ordinary software.
This is not an argument against acting. It is an argument that the tools have to fit a technology with no material to control and, when something goes wrong, often no one to hold responsible. The Atomic Energy Commission worked partly because the thing it watched could be counted. Our harder task is oversight for a danger that cannot.
Kumar Sreekanti Los Altos, Calif.
To the Editor:
Is past really prologue? This editorial has many good points yet fails to distinguish the critical characteristics of A.I. and prior technological creations.
A.I. is not in the same category as nuclear energy and how it was managed by world governments. Its fissionable material is programs, not detectable uranium. It is not based on limited and hard-to-mine ingredients. It’s transportable across the internet and can easily be hidden through the use of encryption. Anyone anywhere can develop and deploy A.I. models or use open-source versions that are freely available today.
Let’s face it, the cat’s already out of the bag. And in the case of A.I., it will take another cat, most likely an army of them, to recognize its own kind and alert us, along with the hardware and systems we use, to its presence — as either friends or foes. We’ll have to rely on A.I., as we do on the body’s immune system, to recognize such pathogens and prevent their destructive acts. It’s much like fighting fire with fire to protect the spreading inferno.
Steve Schaffer Alameda, Calif.
To the Editor:
Tressie McMillan Cottom (“A.I. Is a Threat, but Not in the Way You Think,” column, nytimes.com, Sept. 19) is absolutely correct in pointing out that A.I.’s predicted potential harms have, in many ways, already come to pass. If one extends the framework of her discussion to industrial catastrophes writ large, her argument only gets stronger.
We, as a society, for better or sometimes for worse, have dealt with technologies and their industrial processes inflicting harms on all of us. Nuclear plants have melted down. Toxic chemicals have been released over wide areas. Opioids have been overprescribed, leading to an addiction crisis.
In all these catastrophes, we’ve come to understand that human decision making was the root cause. And in each case, if political pressure had been brought to bear, different outcomes might have been possible.
In this way, the oligarchical push to advance A.I. is just more of the same, and the response, as Dr. McMillan Cottom argues, should not be panic. It should be democratic decision making.
Saul Tannenbaum Boston
To the Editor:
After reading Ezra Klein’s Sept. 20 column, “It Does Not Seem Like a Good Idea to Give A.I. Control of Its Own Frontier,” I can’t help but wonder if concerns about “recursive self-improvement, the process by which A.I.s begin autonomously building and improving new generations of more powerful A.I.s,” relate not only to artificial intelligence but to human intelligence as well.
We humans have come a long way in terms of what we know, and know how to do, and yet our intelligence has produced outcomes that often seem out of alignment with fundamental human values of kindness, fairness and decency.
What we do, and do to one another, would suggest that we’ve lost sight of the common good, or at least struggle mightily in applying that idea to our lived experience.
Is it possible, as we wring our hands about how A.I. might kill us all, that we are looking into the mirror and seeing our own reflection?
Jon Schottland Brattleboro, Vt. The writer is a psychotherapist.
To the Editor:
Re “A.I. Fears Called a Hoax by Trump, but Aides Differ” (front page, Sept. 20):
President Trump thinks the fears that A.I. could lead to mass extinction are a hoax. He wrote on Truth Social last month, “The people who say AI is going to destroy the World … are the same people that said, just a short time ago, that the World would be extinguished by ‘Climate Change.’”
Turning that around, the same people who claim that A.I. is a hoax are the same people who think that climate change is a hoax. And with the huge increase in wildfires, rising sea levels and extreme weather, we can see where that position got us.
It is tragic for the United States and the world that those are the people who are in charge of policy.
Burton Kushner Madison, Wis.
To the Editor:
What is it that feels familiar about President Trump’s latest protestations about talk of the potential catastrophic impact of A.I.? Why do I feel that I’ve read this story before? How come I become more nervous when he says the fears are a hoax? Oh, wait, yeah … what was it called? I’ve got it … Covid.
Stuart Fischman Mission Viejo, Calif.
The post Controlling Risks Posed by A.I. and Other Threats to Humanity appeared first on New York Times.
