Two weeks off from Updating feels like a lifetime, siblings, but I know you’ve been in good hands and I hope those hands have been gentle…I’m sorry, I really don’t know what’s happened here, we’ve wandered off somewhere very odd very quickly.
To bring us back on track may I please ask for your indulgence while I invite any advicetech or fintech providers doing stuff aimed at advisers to consider being part of the second Advicetech Catwalk in June? This is our tech showcase event, it’s great fun, has acted as a good springboard for some firms, and is completely free to take part in. We’ll do all the work, you just come along and show your stuff in front of a couple of hundred advisers and industry folk.
You can find out how to enter – and indeed get tickets – here.
With that punt done, let’s stay on tech (even though there’s lots of regulation we could talk about) and get all AI about it. I had fun the other week being on a panel at the Edinburgh University Business School, organised by CISI, about ethical AI in financial services. Now, I am as you all know a complete buffoon with mush for brains and so I was pretty intimidated to be up against a solicitor specialising in AI, someone who actually writes code for these systems, and one of the world’s leading natural language and AI experts who also happens to be an academic at Edinburgh and co-founder of a fintech business too.
I also don’t really know about ethical AI (you can tell the audience was in for a treat) but what I do know about is our industry, and the way new technologies find their way into adviser businesses in particular. I mean, I know other things too, that’s not the only thing I know – I used to be really good at top slicing, for example. Less so now, but it never leaves you.
A couple of things jumped out at me, and I just offer them here as food for thought as we all get super excited about AI – and of course many of the firms who will present at Catwalk will have some AI inside.
The first is about training. We all know AI is trained on big data sets. But do we know what’s in that set? And even if we do, have we considered what that means? If we train a generative AI on One Gazillion suitability letters, for example, it will learn what’s in them and do a decent job of replicating that. But if those suitability letters have common flaws, biases or tilts built in, what happens? Those flaws are ‘codified’ or hardened into the system and you need to go in with a pair of tweezers to sort it out. There’s a great article on this here from Quietroom. So, AI users: what was your system trained on, and what safeguards exist to ensure biases and tilts aren’t codified into it? I bet you don’t know…
The second was about how firms buy these services. Huge enterprises hire big consultancies and whatnot to help, and they can do the work of digging down into what’s there. Most firms in our sector can’t do that; it’s a case of trusting the nice provider who has come and brought you doughnuts. This is fine while AI is only being used for fun things – marketing emails or whatever – but when you start getting serious with client outcomes, how equipped are firms to make a real quality decision not only on outputs but on how it’s put together?
If you want something that will really bake your noodle (there is no spoon, by the way), one of the big AI developers was on the radio a month or two ago and admitted that they, as the architects of the AI, didn’t know how it worked. How’s that for a punch of reality? It’s already semi-sentient and has started to build its own way of doing things, just like Hex did in the Pratchett books. You as advisers have a duty to avoid foreseeable harms – how can you judge what’s going on when even the experts don’t know?
Progress is as inevitable as the changing of the seasons, and we all have to get used to these new ways of thinking. But if we want to do things right, we’ve got some new challenges to think about too. This isn’t like wondering which amped-up spreadsheet doing cashflow is more fun to use. It’s different from before, maybe different and better, and it matters.
Your music choice this week came out of nowhere when someone played me it – I had forgotten what a great song it is. I watched a few interviews and things with Louise Wener and she remains completely badass and talented and…well, quite a few other things besides. It’s never gonna be this good, so just climb in…brilliant. Please do enjoy repeatedly of Sale Of The Century by Sleeper.
