That Was The Week That Was — 13th September 2026

The experience of watching your garden grow gives you some idea of how future AI systems will feel when observing human life
— James Lovelock, Novacene: The Coming Age of Hyperintelligence

AI was much in the public consciousness this week following the Anthropic's researchers revelations how AI has a 10% chance of destroying humanity by the end of the decade. It wasn;t helped when Geoffrey Hinton commented on BBC's Newsnight that yes, that's probably about right.

A week before I had read James Lovelock's 2019 book Novacene: The Coming Age of Hyperintelligence, which is a much more optimistic take on a possible future with these ever-increasingly intelligent machines. I say optimistic, and that's probably the wrong word as Lovelock believes humanities dominance is coming to to end and it's all part of an evolutionary trajectory where machines will be the top of the intelligent hierarchy, looking at us the same way we now look at plants. Yet he argues they will need us to support the ideas behind his Gaia principle, all of us working together to stop the biggest threat facing all us — that of a planet getting increasingly warmer.

As I use AI in various forms every single day, my relationship with it has changed over the years. I used to subscribe to the idea that LLMs are just giant word guessing machines, a kind of confidence trick ‚ all be it a clever one, but now, the more I have interacted with these systems I've come to see that there is actually reasoning seeming to be going on. It's more than just guessing the next best word.

These daily interactions have made my question the need to learn to code, when I can simply describe what I want and a few minutes later it's made real. Yes, there is back and too between me and machine, correcting mistakes and the like, but the speed this stuff can be made, which allow me to try ideas is seductive. The main thing for me is it shouldn't be completely opaque. Because I know how to code I can query the code it writes, suggests changes, and look at how it's done things, so I can continue to learn. I'm just not doing the typing.

For a long time I've always felt typing code was archaic anyway, which is why I love Node / Graph based systems such as Touch Designer and Houdini. Maybe now this is a perfect coming together of human and machine. The ideas are still mine and I think that's where we are heading. It's going to be more about judgement, taste and direction. We will be more akin to conductors of an orchestra. The conductor may be able to play the piano and read music, but they can't play like a virtuoso. Yet both are needed to create a heart-felt experience.