Alan Buxton’s Post

This year I became a STEM Ambassador. Thanks so much to Simfoni for encouraging us to all do some volunteering. Being one of the STEM Ambassadors is a lot of fun because I get to do coding with kids. What's not to like?!?!? One highlight was being a judge in this Young Coders Competition: the quality of some of the games really was outstanding. Most of the rest of my volunteering is about encouraging kids to get into coding, mainly young girls. Some notable things I've seen these last 6 months: 1. The girl who insisted she wasn't a logical thinker when she was absolutely one of the most logical in the room. Sounded like social conditioning: if so it's heart-breaking. 2. The girl who started off being really bored but then was overheard saying "this is fun". (Yes, yes it is, glad you experienced it) 3. The Stem Ambassador who runs a coding club which he said is 50/50 girls/boys during primary school age (up to age 11) but then has zero girls in after the kids start at secondary school Then there was the kid who asked me advice on how to get promoted in your job. I gave rather a woolly answer at the time. In hindsight I should have used the line I got from a business biography book many years ago and which sums up my philosophy: "Do more than what you're paid for and you'll be paid more for what you do." Obviously I have to mention an interesting #product angle. The STEM Ambassadors website has a really neat dashboard. It tells me that in my 25 hours of volunteering I've engaged with 14 organisations and reach 1245 participants. Makes me want to do more so I can build up my impact. Nice gamification piece.

I've come to realise that my job comes down to answering two questions: 1. Do we have the right people? 2. Are they working on the right things? And phrasing it this way .... well that's basically all management roles!

It was a lot of fun to be a judge for this competition, and being a STEM ambassador more generally. Thoroughly recommend UK techies in my network to get involved.

I do most of my tech learning through side-projects. From things like which AI approach works best for which use case, and what sorts of guardrails you need to productionise it at one extreme, to reminding me how hard software engineering is at the other. A few years back I started a project to figure out how you could create a picture of a company purely from raw text. It was kind of the opposite of what I had worked on at OpenCorporates and a problem that was still not solved (certainly at the time). This weekend I did a bit of debugging to speed up a query. And wrote about it. https://lnkd.in/eYS-sCjT To be fair, this example was pretty trivial, but it still took a few hours and a few dead ends. Perhaps next week I'll do a first stab at a implementing some queuing. Link to my first post about this project: https://lnkd.in/d_tXFG4c

One of the most inspiring tech leaders I've followed, JP Rangaswami, told a story about "how to decommission a mainframe". It's stayed with me ever since. It turns out that the answer is very simple. You turn it off and then tell people 6 weeks later what you did. Then you get responses like "oh, I wondered where that 50 page report that I never looked at went to. Fine, no problem". Of course, if you asked people "is it ok if I decommission the mainframe" then they would suddenly start looking at the 50 page report and get scared about losing it. They would find dozens of reasons why you should keep the mainframe running "just for now, until we can replace the report". We did a similar exercise in Simfoni. We had a bunch of SaaS applications that we were paying for. Ownership of those applications was unclear so we switched a few off. Most of them, no-one noticed. Some of them it turned out were important, so we turned them back on. Especially these days when it's really easy to turn software on or off, we should be spending more time turning things off and seeing what happens.

Mulling over the balance of releasing products early to get momentum and feedback vs making sure you put out a decent vertical slice of your product. And I remembered a case where things were getting a bit time-consuming so we made a conscious decision to compromise on some functionality to get the next release out faster. We thought the compromise was a pretty big deal and something we would have to come back to in a few weeks to sort out and that it would need a fair amount of re-work. But it turned out that the real users saw what we thought was a compromise, but they thought it was a benefit. Just as well we didn't spend all that extra effort on handling all the edge cases and lived with what we thought of as a "compromise". It was anything but a compromise. It was a lucky accident that we hit on a neat feature. It's not a lesson learnt because it's something that we all know in our hearts. It's a lesson reinforced, and something worth reminding ourselves about again and again. Users will always surprise you. If in doubt, get something out.

It's not the AI's that are hallucinating. It's us. We're the ones thinking that what the language model tells us has some meaning. We're the ones who do the double-take and get the weird uncanny feeling when the chatbot writes something that looks good but is not true. The language model is doing what it does. It is creating plausible text based on its training and the prompt that has been fed into it. (Arguably there are plenty of people who do just that, but that's another question). It does raise some really interesting questions about why we equate skills in language with having intelligence. I'd love to read some research on this topic if anyone can suggest a good place to start.

We've all heard that AI is disrupting everything in a completely different way to previous disruptions. For example, see the "Peak Human" vs "Peak Horse" framing. 100 years ago horses became un-needed in much of industry because machines could do what horses could, but better. We had reached "peak horse". The argument goes that since ChatGPT came on the scene in 2022 we are close to when machines can do anything that humans can do, only better, and working people will go the way of working horses. In other words, "peak human". I don't think so. I see this more of an ongoing evolution in who gets disrupted. 20 years ago blue collar workers were seeing their way of life up-ended. Now it's white collar workers being disrupted. Automation is catching up with people who spend their working lives in consulting, tech, marketing etc. Those of us who became used to doing the disruption, and presenting it as "a good thing", are now being faced with disruption ourselves. ChatGPT and its ilk are not the heralds of Peak Human. They are bringing disruption to the disruptors. Which hopefully will help us feel more empathy towards our fellow humans who have already been through this loop. And perhaps we can learn from others who have seen this movie before how best to deal with it.

Can someone help out an aging Victor Meldrew over here. I'm seeing a lot of chatter these days about AI agents. It took me back to my Java days. In particular this book I had: https://lnkd.in/edU5Gd3B In 2001. Have there been some seismic changes in AI agents recently or is this more a case of an ongoing evolution over the past decades that, thanks to large language models, is now having its moment in the limelight? This is a serious question by the way.