From the Mouth of Donald: An 'AI Verification System'
What can such a system do? Can we trust it?
Update: Yesterday, I posted 1800 words on Substack, in which I quoted part of Donald Trump’s speech to the UN. Today, I was astonished to see no reference anywhere to Trump’s announcement that his administration was pioneering an ‘AI verification system’. That was the topic I had written about on Substack.
I had thought it so important, that I placed a type of promoted ‘ad’ on Twitter/X (paying $10), hinting at the topic’s importance, and luring people to read the post on Substack. I couldn’t leave a direct link, so I just pointed the way: saying, in effect, go to my Profile and click the link to my website, where by clicking another link, you will get on Substack and can read the entire 1800 words for free!
However, since studies have proven that computer directions, whether given in Chinese or English, prove equally unhelpful in randomly matched sets of adult Americans [Grok unverified], I am left with substantial doubt whether anyone actually navigated those directions successfully. Therefore, I am publishing the full Substack piece from yesterday here on Twitter/X, (although only premium users will see it in its entirety).
— — —
I’m letting my imagination take the lead on this one… In fact I’m letting it run.
It started with Donald Trump this week going before the UN General Assembly. I turned on the TV while dressing. I looked up and here he was at the podium, bragging about the wonderful buildings he built, with marble on the floor and wooden walls, and at less cost, and on and on. I inwardly gagged; I’m one of those people who happens to experience something visceral at the sound of his voice. I tried switching channels but he was on all of them. I was looking for the on/off button on the clicker when I heard him say [Grok’s verification],
Today, I’m also calling on every nation to join us […] My administration […] will pioneer an ‘AI verification system’ that everyone can trust to ensure no nation is secretly developing these unbelievably dangerous weapons [nuclear and biological].
Hey, that caught my attention. In fact, I had been day dreaming that very topic… (Yeah, that’s the kind of person, I am…) but my daydreams were more in the sense of ‘wouldn’t be nice’ if we had a fool-proof AI system that everyone trusted… even me! Such a verification system would vindicate my optimism about our future as we enter a new age sporting a planet-wide digital nervous system, known by all as the Internet. Yes, having a trusted fool-proof verification (for anything!) would be fantastic. However, beyond ‘nice’, I never took any of these thoughts further.
But my G-d, here was the President of the United States actually going before the world, the f’ing General Assembly of all places, and while he’s caught up in his bragging about these glorious buildings he would build, all monuments to himself — suddenly… like some cantankerous cousin of Balaam’s ass not given the gift of speech — he stopped braying and floated the idea of a trusted AI verification system!
Wow. That was a hell of an introduction to a topic that I had been fantasizing! I, for one, believe something like a trusted AI verification system would be a direct and immediate benefit to mankind. I mean, we got till the end of the century to argue about global warming. But who in their right mind believes we can go on, year after year, without a civilization-ending atomic holocaust. (Here, for experience, I draw upon my year as a nuclear submarine medical officer.) If and when that holocaust happens, I can’t say for sure that our species entirely disappears, goes extinct; but mankind, at most, would not be organized beyond small hunting bands (or terrorist groups). No ‘flux capacitor’ for us; at best, it’s ‘Back to the Past’, and incapacitated for ages to come.
At my stage of life, I see more clearly that our ‘evolution’, i.e., Darwinian, the way we pass down our genes, is controlled largely by our passions. Our nervous system is programmed to leave coupling up to the whims of our limbic systems, the neurological driver of our emotions, our primitive reptilian brain, with little in the way of input from higher, frontal lobe centers. I’m not complaining. It’s probably appropriate and life affirming.
But what do we do on the international stage? How can we get our intelligence to play its part? What can help us avoid destroying ourselves if we use the same emotional brain used to propagate ourselves? Look at our recent wars; are they not also the result of irrational emotions from that same limbic system — whether it’s something pathological like a suicidal psychosis (Hamas), or an ideology that says this is how we carry out the will of our current god (the Mullahs in Iran), or this is how we recapture the glory of an empire past (Putin in Russia)?
After hearing Trump’s idea for a trustworthy ‘AI verification system’, I immediately began sketching out how that would work. As I said, that’s the kind of guy I am. I’m really not ‘smart’; but I do observe things. And I do have my uncommon capabilities, perhaps I’m what today RFK would call ‘autistic’, (although I grew up before Tylenol). Besides an aversion to Donald’s voice, I do have other unusual ‘sensitivities’. Perhaps they come out as a mind-set that produces uncommon patterns, like outliers. I never met a doctor who designed and wrote their own software platform for clinical practice; mine was called ‘AmigoDoc’. (RIP). At one time there were easily a dozen psychiatrists across the nation who looked to me for software maintenance.
So I set to work. Okay, how would I design an ‘AI verification system’?
I visualized the problem thusly. My opponent wants to secretly build a *nuclear* weapon and my job is to sound the alarm if he actually starts to try. This alarm has to be as tamper-proof and as reliable as possible. I, and my friends who support me, need this alarm to be 100% trustworthy. (I am focusing on the nuclear threat because it already exists and is the most lethal by far. A nuclear bomb gives no time to develop a reactive defense, as opposed to both a mutant virus and global warming. In the face of an atomic holocaust, and having little beyond my childhood’s ‘duck and cover’, mankind would be as vulnerable and as naked as dinosaurs 66 million years ago when that large space rock splashed into (and obliterated) the Yucatán, [Grok’s verification].
The nuclear process we are guarding against leaves clues that are straight forward. You cannot make a nuclear weapon without using materials that are already very unstable and radioactive. Radioactivity can simply and reliably be sampled and detected in any environment. If you find enough quantity, that’s it; slam dunk, the alarm sounds. What you have to guard against is making sure nothing steps between the radioactive particle and resulting message that triggers the alarm. In other words, you must be certain that nothing can step into the path of the triggered message and block its transmission to the alarm.
Think of the ‘AI verification system’ as a microscopic unobtrusive black box that can be placed anywhere, or scattered randomly, or even placed in a button. Now scale up that black box and think of it as a house you want to guard and keep secure. Every possible entrance (doors, windows, chimneys and vents) would have devices to keep out intruders. (The AI part would be used to identify possible suspects loitering outside the house, for special monitoring.) Each entrance would have a different variety of locks, detectors and alarms. In a practical sense, no house can be hermetically sealed. In a theoretical sense, no software in use can be certified as 100% ‘hack proof ’, and computer theory says it’s possible to break into any software in use, [G.v.]. (Even so, your devices, while not fool proof, may still make the cost so high, and so conspicuous in other ways, that alternate alarms are triggered; or so impracticable that the effort exhausts a country’s entire computing power, unable to be sustained or hidden.)
There’s another reason why this software may send a signal that’s 100% reliable and still conform to the theory that everything potentially can be hacked. Trump’s black box will require no input rather than a radioactive particle bouncing off its skin. Otherwise, it’s passive but checking its functioning continually and randomly; try messing with its mechanisms and alarms go off. Because everything’s internal; always under scrutiny, it can be 100% trustworthy. The hacker has no key to copy, counterfeit, or spoof. There is no one to impersonate.
I’m sure computer labs in the West and in Israel have already developed this software, probably many years ago, now tested and refined. It only needs the press and pressure of a strong political will to become operational. Trump’s ‘AI verification system’ will require an ironclad agreement from a forum of countries. I don’t think this will be an issue for countries that are overwhelmingly composed of ‘democracy loving peoples’, (to use a label supplied by Thomas Friedman on Tuesday’s NYT’s lead editorial). This would mean the US, Europe and Israel; i.e., ‘the West’.
And why not; why should they not promise to reveal their ‘secret’ attempts to develop nuclear weapons when either they already have them, or they’re covered under a treaty-bound umbrella by an ally. I presume most Western nations, even Israel, would make an exception to any issue of absolute sovereignty, if they knew that *all* countries in the forum agreed to this minimum standard of neighborly trust (and verify).
No one would be forced to sign on and have these devices implanted into selected sites. However, part of the forum’s *Magna Carta* would be that any nation electing *not* to agree to this AI augmented alarm system would suffer diplomatically and economically. Such countries would be shut out of the world’s commerce and communication systems. Besides the West, I presume that the vast majority of nations would agree to join; perhaps Iran and Yemen would refuse; I would not know about Russia and China.
All the nations installing this alarm system would undergo a specified period of formal installation, with testing and debugging on site; and then the system would go active. After that, any alarm would trigger an immediate process where inspectors would go in and recheck all mechanisms and verify the alarm. If verified, the alarm would then kick off a formal investigation; and at the same time, place all international trade payments in escrow. Any country refusing to cooperate would be shunned, excommunicated and cut off — shut out of the earth’s international internet and economic ‘nervous system’, and treated like an international leper.
Something like this has always existed as a utopian vision, and always seemed just as unrealistic. However, today, we have an Internet that functions like a planet wide nervous system, augmented with satellites and AI. Suddenly this vision becomes a real possibility.
Wouldn’t it be ironic if something like this, proposed and pushed by Donald Trump, actually becomes a reality? And could not the international community build upon this reality, and reduce the possibility of destructive war? Would it not be appropriate to give Trump the Nobel Peace prize for introducing this idea and providing the political will power needed to bring it to fruition?
Think, kiddos, if it comes about… Not by Churchill, Lincoln, or Socrates, but by Trump. I’ve reached an age where I find that there’s room for the intervention of a divine Providence, by whatever label you care to use. I’m happy with HaShem, the Hebrew word for G-d, (with the dash built in). How could something be established like an ‘AI verification system’, world-wide, by Donald Trump without the hand of Providence? That’s the way I would see it.
I hope it really does come about. Like AI, I expect it to be open and available to all.
I’ll be ready to trust it.
— — —
Afterword: If you’ve plowed through these 1800 words, how about leaving a comment?
Even better, while you’re at it, how about subscribing? That’s a sure positive feedback. Your subscription is absolutely free and will always remain free.
Right now, I have about a hundred subscribers. At some point, I would like to start charging for additional subscriptions; but coming in early, you’ll always be free. There’s a red button below that says “Subscribe now”. You’re welcome to pledge something, but “No Pledge” is perfectly fine. (It’s healthy to be suspicious on the Internet. “No pledge” guarantees it’s free.)
Here’s the ‘Subscribe now’ button I told you about. I hope you’ll click it.
from ALAN JACOBS
Arnie,
I kind of promised myself, I wouldn’t get caught up in responding to your Substack emails. But I did a speed read of this while sitting on my ElAl flight back from my abbreviated trip to Israel. I just had the inclination to give you a little reality check. Maimonides, the Rambam, said basically, don’t look at the man who is saying it; look at what he says. No question Trump is a megalomaniac bore, but he’s a lot of guts and comes up with good ideas. You just need to get over his personality and filter out the bullshit. I strongly believe that he knows it’s bullshit or he gets a kick out of doing it anyway. One of the most important things you can say about him is that he’s clearly the most transparent politician ever. But that’s all just basically an aside. Oh, one other thing; you should refrain from quoting anything from Thomas Friedman and probably from the New York Times as well. Neither of them could pass the simplest polygraph test. Trump bullshits but they lie intentionally.
Except for Israel and a handful of a few small, pretty much inconsequential countries, and maybe the United States, there are no democratically inclined or morally decent countries in the world. Countries would just find a way to subvert your notions and present as truth their own. All you have to do to know that's correct is to look at the statements coming out of England or France - let alone Spain, Norway, Iceland and other so-called democratic countries.. It would be great if there could be a system as you proposed, but it ain’t gonna happen.
Remember, I love you and this has been dictated and unedited, so if anything seems unclear, do your best to figure it out.
Alan
Another well-writtten and thought provoking daydream from Dr. Rosen. The flaw is thinking that Trump would not figure out a way to use it to his advantage. He would literally hold a skeleton key with enough leverage to cause untold damage. He might trust grok,but I can't identify an individual or organization with unlimited power. Look at the mess Donald (and others) have made of the world our children will inherit.