
This isn’t an AI newsletter per se, in the sense that I don’t always write about it. That said, however, I do write about it fairly often, mostly because I don’t think there’s anything else happening right now — apart from maybe crypto — that blends the surprising and the terrifying and the confusing and the potentially evil so perfectly as AI. That’s the Torment Nexus sweet spot! (You can find out why I called the newsletter that in this post, in case you don’t know the story already). Is the kind of artificial intelligence — or whatever you want to call it — that we see all around us now an incredible technological advancement? No doubt about it. Regardless of what you think of AI’s current abilities or potential, it’s still mind-boggling to think of how far we have come in the three years since ChatGPT and other tools first appeared on the scene. Are they intelligent in any real sense of that word? Sure. Are they conscious? Who knows. Is AI an unalloyed good? Of course not. Does it spell doom for mankind as we know it? Maybe, but probably not.
I’m not here to take sides in the “Is AI Good Or Evil” debate, to be honest. There are people much smarter than me who already have both sides of that covered, and most of them (although not all) have a much deeper understanding of the technology and its limits than I do. In fact, one of the things I find so fascinating about AI right now is that there is so much disagreement even within the field itself, and even among those who helped create the technology we are currently using, like former University of Toronto professor and former Google AI staffer Geoffrey Hinton and Meta chief technologist Yann LeCun and McGill University lecturer Yoshua Bengio. Are we close to AGI? Geoff says yes, Yann says no (and prominent AI critic Gary Marcus says hell no). Does AI pose a mortal danger to humanity as we know it? Yoshua and Geoff both say yes, Yann says no. I’ve written about this before, and also about the question of AI and consciousness.
In the same vein, I was interested to see two recent studies of AI that seemed to point in completely opposite directions. In one, published by Apple’s Machine Learning Research project and titled The Illusion of Thinking, scientists raised some significant doubts about AI’s “intelligence,” pointing out that even the latest more sophisticated AI engines from companies like OpenAI and Anthropic and DeepSeek couldn’t solve — or took much longer than they should have to solve — a puzzle that an eight-year-old could probably figure out without too much trouble (a block and peg puzzle known as the Tower of Hanoi). They seemed to have a tough time with some other simple puzzles as well, including the “river crossing” puzzle, in which the test subject has to get three conflicting objects (fox, chicken, and bag of grain) from one bank to another even though they can only take two at a time. In fact, the AI engines had difficulty even after the researchers gave them clues that pointed towards the solution! Here’s a summary of the paper:
Note: This is a version of my Torment Nexus newsletter, which I send out via Ghost, the open-source publishing platform. You can see other issues and sign up here.
Continue reading “Is AI smarter than we are or stupider than we are? Yes”




























