The Sweet Embrace of Cold Logic

AI in another setting

I am not organised by nature. Or rather, I am organised in independent microspheres of certainty surrounded by a vast frothy ocean of WTF. I know exactly where I put my wallet, keys and phone each night, and in fact all the keys in our house now have AirTags on them. Even the dog has an AirTag, because she’s a mad-eyed fruit loop with no sense of direction and a desperate desire to follow Mrs Harkaway wherever she goes.

But my office - despite my desire for one of those sheer, beautiful workspaces that combines driftwood flooring with stark modern book cases and a single aluminium computer at a standing desk - looks like a cross between a junk room and … no, actually, it just looks like a junk room. Curtains I haven’t put up, crates of my own books, a step ladder which has no other home, the boxing from my daughter’s school laptop, my folding green screen - in use as a sun shade so I can see my computer screen in the early mornings - an old Anki Overdrive set kindly donated by my brother which the kids have not taken to, and inevitably a file of old papers which I haven’t actually filed so much as vertically ignored, and which have now slithered all over the carpet.

This is the nerve centre of the rather intense operation that is my professional and personal life, and honestly it’s not conducive to smooth function. I used to claim that the serendipity of it produced more interesting work, but I’ve reached the point where I actually just don’t have time to fix it.

Enter the AI in another form. My Instagram feed has filled up of recent days with ads for Motion, the AI To-Do list. This is not Carrot, the hilariously murderous HAL of To Do list apps, which I rather liked but which in the end wasn’t quite funny enough and instead made me feel genuinely attacked. This is an actual algo self-management system which supposedly nudges and tweaks your schedule to get more done in less time and help you stay on top of things. Although the bumph is mostly about team management, the first question it asks you when you click on the seven day free trial is whether you’re one person or many. I have selected “individual” although actually I function in various loose knit unofficial teams, all of which probably drag slightly because they have me in them. What I haven’t done yet is actually start the trial because I’m not sure I have time to onboard myself and work out how to use this thing. Seven days seems like a frankly piffling amount of time to bootstrap my understanding of a new app and work out if I like it.

One of Motion’s selling points is “channels” - basically subheads for your life. That seemed like the kind of thing that would take me a while to come to terms with, so I sat down and wrote on a Post-It note what my channels would be. Well, I say a Post-It. It turns out there are quite a lot of channels and some of them have sub-channels. In fact there are dozens of the fuckers. It’s possible that I sometimes feel overwhelmed because I’m overwhelmed. The Post-Its now occupy the space to the north of my mouse, just beside my Yeti Nano; if anything could have persuaded me more thoroughly that Motion might be worth trying than this exercise in why it wasn’t worth trying, I don’t know what it could be.

This seems to me to be the beginning of a space where AI of the kind we’re developing - statistical AI - has a real place in the world. If it can actually take the organisational load off and intervene in the moments of norn-like halting state which afflict me when I’ve got too much in my in-tray, that’s a huge benefit. The only thing is trust.

Yeah.

I have to trust the system to deliver the right answer at the right time. That’s fine. It’s unlikely to do worse at that than I do, and I can always go off-script. This isn’t a self-driving car; reaction time is not an issue.

But trust more broadly… harder. I have to give the system access to my schedule and my priorities, and I need to know the end point of propagation of that info. Does it get used, “anonymised”, to develop the AI? Does that package then get sold on as behavioural data?

When I did the BBC Archive Hour show (I’d link it but it doesn’t come up in my search, and if you find it PLEASE ignore my horrible maths in the camera per square foot question - I just ballsed it up) I interviewed a few people about their attitudes to smart homes and surveillance, and the general burden of their song was that they were fine with their data travelling a certain distance, but unconsciously assumed a kind of exhaustion of momentum which doesn’t really exist in the digital world. If you roll a ball in real life it comes to a halt, and they were assuming that - say - the data from their smart toothbrush stopped at their dentist’s office. But it doesn’t.

I want the help, god knows, but I really don’t want to train another system to manipulate me better. Cryptocurrency was about trustless architecture. Perhaps we need trustless AI, whatever that means.

Anyway, when I set the ball rolling on Motion - if I do - I’ll tell you how it goes.