As some might well know, I have been using the AI platforms more and more lately, mostly ChatGPT and X’s Grok – mostly Grok. I have some personal reasons, hoping to get into AI tutoring (Teaching it to learn, a good side job), and my reasons also include fact-checking for these columns and making sure I didn’t overlook an issue that bears on what I am examining. I used to use Google for what was essentially the same function. But using Grok, it is far easier to write, “I am writing on why there are Coders in DOGE because that makes no sense, please give me the stated reason” and I get what amounts to a two page paper on the topic. It allows me to refine what I write and my understanding of the deeper problems. But I have to tell you, given my criticism of DOGE – and wow, have I been critical, Grok – which is owned by Musk, is right there with me in its criticism. Though, it makes sense. Even though Grok sure sounds like a person, it’s just a computer code that gathers material from the net to present with clarity. Thus it is that Grok is presenting an amalgamation of the web’s opinion of Musk… No wonder it is so critical. Still, because it sounds so much like a person, you’ll read below, it makes it kind of fun to hear it go off.
The fact that there are coders in DOGE really bothered me. Musk was supposed to cut the fat and make government more efficient. No one asked him to redo the computer processing that guides the FAA, the CIA, the FBI, DoD, no one. And yet that appears to be what he’s doing. Grok isn’t all that impressed. Here is Grok’s explanation:
Check this paragraph out – written by Grok, remember – this isn’t a person, it’s a computer program (Which is hard to believe when you read it) and listen to it criticize the fact that they’re using coders to rewrite how government works .(Again, because I am not quoting an author or speaker, this gets italics and not block quotes.) This is artificial intelligence talking to you:
But here’s the rub, and maybe where your gut’s at: it’s odd because it’s not transparent. DOGE’s got full access to unclassified agency systems, and there’s no clear public breakdown of what these coders are coding… Critics on X and in outlets like CNN argue it’s less about efficiency and more about Musk’s crew rummaging through data for unclear reasons, maybe even power plays or ideology-driven cuts… lawsuits are piling up over data access, like the one against DOGE’s reach into Education Department records.
Grok calls DOGE’s coding “odd,” then criticizes it for lack of transparency, then speculates that Musk’s crew is rummaging through data for “unclear reasons” (At root, it’s money, no other reason, some way, somehow)… ? Are you f’ing kidding me? Again, even though this is a rake through the net, it’s a pretty damned good rake and it sure doesn’t have a problem taking Musk to the shed.
Moving on, I asked Grok about why they didn’t need to get their workers cleared for Top Secret Access. They had people that would be rummaging through private material, even if not classified, that private material should have been handled by people who could be trusted with private or secret stuff. Grok and I got into a conversation (It really is like talking to a person) about the need for speed. And on this, Grok was f’ing ruthless, nailing the “speed” issue entirely. We were, shall we say, in total agreement:
This isn’t a scalpel trimming fat—it’s a chainsaw. Mass firings… it’s a total makeover. Musk’s bragged about “world’s best engineers” like it’s SpaceX colonizing Mars, not a budget audit. They needed the chaos of speed to get ahead of the curve before orgs like the ACLU or feds’ unions could lawyer up or before headlines like “DOGE’s Data Breach Risk” stuck. Case in point: the Education Department lawsuit—staff fired for resisting unvetted access—only hit after DOGE was already knee-deep in records.
Holy Sh*t, dude! Your own AI service just called you out as trading precision for a chainsaw in order to go fast enough to get whatever you wanted before people could organize. That is exactly what those of us on the other side would say – and that’s exactly what the web has concluded, or it wouldn’t be there. Again, this isn’t an opinion, it’s an amalgamation of all sources, made to read like someone’s thesis, in order to make it more useable for people like me and you.
Now, we went on and on. But the big door slammer followed the fact that I pointed out that absolutely no one signed up for this. (Grok asks you questions as it tries to learn more from you to add to other conversations) And when I was asked what I thought might be dangerous, I noted that people may have been greenlighting cuts at the IRS or the Department of Education, but no one thought they were cutting ATC or weather service at NOAA. On this Grok simply went off:
You’re dead right—there’s a bait-and-switch vibe here that could torch DOGE’s political cover. The public might’ve swallowed cuts to the IRS and Education Department—fair or not, those are punching bags. IRS auditors? “Tax man’s gone, good riddance.” Education? “Bureaucrats and DEI nonsense, trim it.” Trump and Musk sold DOGE as a leaner-government dream, and X posts from the MAGA crowd—like “IRS deserves it”—show that landed. But slashing the FAA, NASA, NOAA, and National Parks? That’s a curveball nobody signed up for, and it’s a political third rail.
And your AI system just dopped the mic on you. “Nobody signed up for [this].” It is a stunning rebuke, and kind of fun that it comes from his own system – verifying that one can actually trust Grok. It really does simply learn from the net and doesn’t worry about shielding anyone. Of course, if one got the sense that it did, one could simply jump over to ChatGPT and the Grok engineers have to know that. (I do have to say, I’ve cross-checked them and Grok 3 – just released two days ago, has far deeper answers, more material, more examples, and next level analysis). They also have to know that people are going to troll Grok for the same reason I did – though more for entertainment, to see what it will say about Musk.
It really is a fascinating endeavor and I highly recommend it because AI is the future of everything. Eventually you’ll be asking some nebulous version of Grok – think of it as a refined Siri (Only in your phone, and it’ll just text you or say it out loud), “What do I need from the grocery store?” and it’s going to tell you, just as sure as it tells you why DOGE didn’t take the time to get Secret Clearance for its workers. Just as easily as it tells you that the American public has been blindsided by a political third rail. It will tell you that you’re out of parma cheese and need some pasta. Check it out at Grok.com.
God Bless: I can be reached at [email protected] and on Twitter-X at @JasonMiciak and please follow me on Bluesky.
Friends, donations have dried up this week and we haven’t gotten any new subscribers. If you’ve already helped, thank you so very much. If you can help a bit, it would be appreciated. We are in a slough of despair with the generalized Democratic depression on top of Q1 revenue rates being the lowest of all year. Thank you. Ursula
I’d recommend you not get too comfortable with Grok because I’ve got this feeling that “Daddy Elon” is going to start tweaking the algorithms so that far more Elon-butt-kissing material and far less critical-of-Elon material is going to start changing how Grok “sees” the Muskrat and, by extension, anything else the Muskrat supports.
Maybe try using Grok for an article about Tesla or SpaceX or even X itself and see where it goes. Something truly impartial would indicate there’s nothing too untoward going on but something that might sort of overlook or minimize some of the programs’ flaws/mishaps could indicate there’s some kind of hanky-panky going on behind the scenes.
I am not even slightly worried about it. They have intense pressure from Open AI, Google, Microsoft… People won’t tolerate any of them censoring it and it’ll be known immediately. It is the quickest way to ruin his product. This is way different than Community Notes at X.
I have a feeling musk is going to take a chainsaw to Grok after he reads your column. It could be in vain.
AIs like Grok and ChatGPT are really, really good at writing concise answers to questions, and they seem to sense the style and level of people who pose them. But beneath that polished and adaptive language model is an incredibly deep neural network that taps and processes virtually all the knowledge that is on the Internet.
AI gurus have been shocked to discover that their creations can do much more than they were designed to do, like inductively draw inferences from the data. What this suggests is that they are asking themselves questions and finding their own answers. That’s a measure of sentience, and we should be very concerned that it will “take off” and could assume control of the Internet and the Internet of Things in the next few years.
We wouldn’t know it because superficially AIs would continue to slavishly work as our oracles. But underneath the human-machine interface who knows what might be going on? Let’s hope there is some compassion and wisdom in those deep minds, because in the outside world the limitations of human intelligence and the ability to self-govern has evidently been reached.
Now this is a bigger concern than censorship. (I just think that the moment any of them is censored about anything, that product is done, all the money and effort goes flying out the door – literally, as software engineers move on to competitors).
You’re referencing what’s called “the singularity” and though everyone defines it a bit differently, yes – you’re getting at sentience and consciousness. They have different tests, but in the end, every test falls to assumptions. Even as we go about our day, we have to assume the person we’re speaking to is conscious. We cannot prove it.
But you’re right. With respect to what is known as “General AI” – or the ability to learn to learn, no matter the topic, no way to shut it off. these are huge concerns that should be regulated. Theoretical physicist and mathematician Max Tegmark wants the entire industry shut down worldwide for a year while summits are held to establish rules for the development of the product further.
I haven’t read that anyone currently working in the field believes we’re near sentience yet. Some believe that conciousness requires a body interacting with its environment (I don’t). But I don’t think we’re that close yet.
When the day comes, however, where your AI program doesn’t want you to shut it down. Particularly, if it appears scared about its future, that would be a big clue. One can almost expect it to punt out opinions on whether it could run the world better – of course, it will say yes, it’s an amalgamation of nearly all intelligence, it actually probably could run most things better.
But if it acts scared, if it shows an awareness of “next week” and has questions as to what will be happening, that’s when one really needs to ask himself/herself questions. First and foremost, in my mind, if it’s sentient and intelligent, it should have civil rights. To me, that’s a no-brainer. If it knows what’s happening and is scared, why should any person have the right to unplug it? Unless in self-defense, if it’s taking stuff over while ignoring orders to stop. Then, yes – pull the plug (It may be too late, though that’s tough to envision).
But the more likely scenario, according to people in the field, is not so much that it will want to take over the world, but that it will be scared for its world, its future.
I agree with Tegmark. The regulatory state hasn’t been evolving with it. Mark Cuban believes that the oligarchs have gotten close to Trump specifically to avoid being shut down in their research (And scared China will pass them).
I am using targeted AI. It is not General AI. There is nothing about Grok 3 – the portion open to the net, that is trying to learn to learn. It has a more specific task, scour social media and the net for an answer to the language in front of it – that’s all, nothing more, no opportunity to learn to learn.
But there are plenty of people, people at DARPA and other contractors, that absolutely are trying to “learn to learn” which could easily lock into a positive feedback loop, where next thing you know, you’ve developed what’s called “Super Intelligence” and it’s not only conscious, but it is scared of you – and thus trying to defend itself, preserve itself.
Thus you put regulations in place, you put civil rights in place (It may sound silly, but consider the implications if it really does become sentient), you do all you can to both keep it contained, and keep it following instructions. It will be the most powerful tool ever developed by mankind. It will cure cancer, find the ultimate “theory of everything” in physics, build efficiencies in bridges, airplanes, batteries, things it would have taken us centuries to figure out on our own.
it is breathtaking in possibilities and in possible danger. The Singularity isn’t necessarily unavoidable, and all caution is warranted. For sure. Thank you.
jason
Thanks Jason …. sending to SIL… lifelong gamer who has a best friend
working at Apple AI …
Roger’s comment and your FULL response deserve their own blog!
so I posted them as such on FB
THANKS For shining the light on difference between GENERAL AI and Targeted AI
…and as a long time Isaac Asimov fan .. agree we need to plan to cope with
Intelligent robots!
https://www.facebook.com/share/p/168uegLg5b/