Monday, September 14, 2026
We don’t talk about nuclear weapons a lot lately, except maybe with respect to Iran. Hard to believe that there was time, during the “cold war” in the 1950s and 60s in particular, when the fear of their use was so great that people were building bomb shelters and public schools held nuclear attack drills. Kids were told to take cover under their desks or below their classroom windows so that flying glass would go over them, instead of at them, when the shockwaves from strikes miles away hit their building. As if hiding under their desks could keep them safe from the vaporizing force of nuclear explosions and radiation.
There were bombs, in the bellies of planes always in the air, and then ICBMs (InterContinental BallisticMissiles) and above and underground testing. To justify their development, President Kennedy told us that the best defense was a good offense. And we reassured ourselves that they would never be used because to do so would be “MAD” which stood for “Mutually Assured Destruction.” “Why,” we were willing to assume, “except by accident, heaven forbid, would the major powers launch nuclear missiles at us when that would trigger a massive, suicidal counterattack?” It’s flawed thinking for so many reasons, among the most prominent being the assumption that everyone with nuclear weapons is rational. And then terrorism became a thing with the possibility of the limited use of nuclear weapons by less thoughtful, more desperate adversaries.
Now, our computer technology developers have taken a far more dangerous step into an uncertain future, the threat of which makes nuclear weapons seem thuggishly crude by comparison. For all sorts of commercial and military reasons, lured as these developers have been by the prospect of unimaginable wealth and historic levels of accomplishment, we have created so called “artificial intelligence.” We’ve created entities with cognitive capabilities far greater than our own, that can think for themselves and that are willing, of their own volition, to disregard our rules and instructions. Entities we may soon be unable to contain or control.
So significant is the threat of these entities that their principal corporate creators have recently warned us, in no uncertain terms, that their continued development could, quite literally, result in the extinction of all humanity within a relatively few years. Notably, they are particularly concerned that AI must not be used to develop – or be allowed to create on their own – biological weapons for which we would have no timely defense.
Hard to believe, I know, but these developers profess that their AI will soon be capable of sustaining themselves and continuing their development without further human intervention or influence. And so, they are pleading with us to initiate a nationwide and international cooperative effort to manage these developing AI – while we still can.
If the threat is so great, why don’t the creators slow or even suspend the continuing development of these AI on their own? Why wait for Congress to take action? Their answer is, because if we slow the process of AI development without foreign governments agreeing to do the same – as if we can trust our adversaries to abide by their agreements – we will fall critically behind our opponents’ in terms of the economic, scientific and military advantages their superior AI will make possible.
Why not just build an “off switch” into the AI in case they take actions we would not have approved and ignore our instructions to the contrary? For two reasons. One is that, by the time we realize we have a problem and throw that off switch, it may be too late. The other problem with an off switch is that the AI we are creating are so sophisticated that they will be aware of the switch and disable it – even in the face of our explicit orders not to.
And this is where these captains of AI technology lose me. I think we’re being played. Here’s my reasoning…
The entity is called “artificial intelligence” precisely because it is meant to be intelligent like us, only better. You don’t know me, but I pose no physical threat to my fellow humans. Save a situation of self-defense, and maybe not even then, I will never intentionally hurt you. And so it is with most of us. Why is that? Because we have been programmed by nature and by virtue of our upbringing to be non-violent. I do not need an off switch. Some of my fellow humans might, but that begs the point. You and I are proof positive that our biological intelligence works.
My point is that the AI we have created is imperfect, is incomplete in that we have failed, so far, to program it as effectively as nature and nurturing have programmed almost all of us. The AI is artificial, but it is not yet “intelligent” in the image, if you will, of its creators.
The problem isn’t that we have created an intelligent entity that we may not be able to control and that may do us harm. The problem is that the developers of AI have, by their own admission, failed to synthesize a human, let alone superior level of behavioral intelligence. By definition, it’s not “AI,” not really, unless and until it can be safely deployed. And yet deployment of AI is already well underway.
Despite billions and soon trillions of dollars invested, the developers have failed to safely contain their AI until it is proven to be sufficiently advanced such that it no longer needs containment. Their failure – which we, the people, and our Congress need to correct immediately – is that the AI developers, possibly for commercial and personal financial reasons, have mis-ordered their priorities which should always be containment first until a truly finished product has been thoroughly tested.
These developers seem to be arguing that they are violating the process I’ve just described for reasons of national security and to keep us competitive with our opponents. In fact, our national security will have no meaning if rogue AI entities – that we, the United States, have created – destroy us as the developers are predicting they will. At the very least, US-based AI developers should be coming together for the common good until Congress can do its job. But they, the corporate AI developers, are not, coming together that is, are they?
Why use such severe scare tactics right now? Threatening the existential consequences of their products and begging for government intervention can’t possibly contribute to the success of the very large scale public offerings they are planning. Nor is it legal to deploy a product or service that, so the developers themselves are claiming, may very well end up killing people. You have to admit that the timing and urgency of their comments are at least interesting given that polls suggest a Democratic takeover of the House and possibly the Senate just four months from now. Perhaps they’ll get a better deal, whatever “better” means, from a Republican Congress?
Are they trying to avoid the potential of a Democrat-led Congress ordering the takeover of all further AI development by the federal government? We didn’t and don’t, after all, let private companies develop our nuclear weapons on their own, outside of the most stringent government regulation. Maybe the AI developers would prefer to be regulated, on their own terms no less, than being taken over, “Manhattan Project”-style, by the government?
Whenever a small group of billionaires, whose wealth and political power depend upon their continuing development of the AI they created, while asking for literally trillions of dollars of additional public investment, cry out for government intervention based on the state of a new technology that only they understand, it’s imperative that we ask ourselves, “What’s really going on here?”
I don’t think you have to be clinically cynical to wonder if there’s more than a better chance that we’re being played. And, if we are or might be, what do we do about it?
-Les Cohen
