I expected to feel more on top of things by now. Six months into studying Transformative AI at PolyU, the honest summary is: I know more, and I’m more confused. I’ve decided this is probably a good sign.
The volume of material in the first semester was significant — 46 papers and documents covering everything from the neuromechanics of fruit fly locomotion to the geopolitics of Chinese AI chip development to red-team studies of autonomous agents. I read a lot. Some of it I understood deeply. Some of it I understood just well enough to know what I didn’t understand.
What shifted most wasn’t specific knowledge — it was the frame. I came in thinking about AI as a technology to be managed. I’m leaving the first semester thinking about AI as an environment to be navigated. That sounds like a subtle distinction but it changes almost everything about how you approach the strategic and governance questions.
A few things genuinely surprised me. The connectomics research — uploading a biological brain and watching behaviour emerge from structure alone — was not something I expected to find relevant to business strategy. It is. The CooperBench finding that AI agents perform worse when working together than alone — that stopped me in my reading and made me sit with it for a while. The Claude Mythos System Card, describing Anthropic’s decision not to release their most capable model — I hadn’t expected to find a large AI company exercising that kind of restraint, publicly, with full documentation.
The semester did what good education is supposed to do: it gave me better questions, not just better answers. I’m working on the answers. Ask me again in a year.