Weeknotes

Emily Macaulay

Emily Macaulay

Emily Macaulay

What I’ve been doing

Adam left our team last month. Obviously this is sad for us as he was great at what he did and a nice energy to have around, but he has made his part time role with TUC a full time one and frankly that’s great for everyone the TUC helps. There’s some administrative processes around someone leaving - as they prepare to leave (including a leaver’s discussion to capture their reflections on organisational culture) and once they’re gone (such as shutting down accounts and deleting data). So I’ve been doing that. It now means there’s only three of us in the team, and that’s a 2.0 FTE at the moment. As a result I’ve jiggled our meetings a bit and rolled all our checkins into Confab - that’s now taking place weekly again on a Monday morning. It’s about being adaptive to the needs of the organisation.

I promised (in my last weeknotes) a selfie with a bus stop if I found one of the “Let’s Talk AI” ones…I found one, so here it is.

Selfie of Emily standing in front of a bus stop

I’ve been trying to finalise the annual report (my yak shaving experience was unsuccessful which I’ve found hard to accept and had to send a “confession” email to Jeni explaining that I couldn’t do the annual report upload); and we’ve had the draft end of year accounts come through to review. There’s been the usual end / start of month activities such as monthly reconciliation.

Since my last writing there’s been a decision that we - with others - will be running a Expo Booth at the AI For Good Summit in Geneva in early July. I’m trying to wrangle and drive the logistics on that. Around the Booth itself we’re trying to support people booking travel and accommodation, and securing funding for that. We’ve never done an Expo before, and doing one as a collective means a number of different voices/opinions. It is very busy! I am incredibly lucky to have been asked to attend and staff the stand for the week though, so … I’m off to Switzerland! That’s exciting and hopefully I’ll have a couple of hours each evening to stroll and explore the city too.

What I need to take care of

There’s still lots to do for the Expo so that’s my priority.

Getting the end of year accounts finalised next week will be important to then get everything published and re-forecast some elements of the budget for this year without Adam’s time included.

I really really need to make some progress on the regeneration options for Connected by Data that we need to think through, review and iterate - all ahead of a Board meeting in September where a decision will be made. September is rapidly approaching. I’m struggling a little to find the focus / headspace chunks of time to do that work.

I’ve got some associated actions around that too. Writing an ‘ending well’ plan…although as I note that here I realise that will best be done after the Board have made their decision in September as they will have an impact even though whatever they decide it won’t be Connected by Data as it is now. I also have a couple of reflective blogs to share too. Memo to me: Create a ‘project’ called ‘Ending Well’ on the website to tag related blogs etc. [Edit: I’ve now done this]

We’re also in the early stages of planning a public sector focused event in mid/late October; a Green Party Conference associated event (2-4 October); and an input to the TUC Annual Congress (13-16 September). It certainly won’t be a quiet summer!

What I’ve been inspired or challenged or moved by

In some weeknotes in February I mentioned how I’ve been following the weekly Premier League football match predictions by the usual pundit (Chris Sutton), a changing celebrity guest, and - for the first time this season - Co-pilot (LLM). As we’ve now finished the season here is the final result.

Going into the final day of games Co-Pilot and Chris Sutton were level on the number of weeks where they had an “outright win” (more points from their correct predictions than anyone else). But Chris Sutton was in second place as he had less weeks (2, to Co-Pilot’s 4) where he had tied with someone else - so he needed to score more points in the final round of predictions than Co-Pilot in order to win.

He didn’t.

The human predictor only managed two correct results with no exact scores, giving him 20 points. Co-Pilot managed four correct scores and no exact results, to end up on 40 points.

Screenshot of BBC website showing final standings

380 Premier League games were played this season and the result is the over the course of that Co-Pilot, using the predictive LLM power that it is, was more accurate than an expert human. I wonder what best explains this. There could be something about a human being more likely to factor in context elements (a new manager, something happening off the field that’s upset the players etc) and perhaps this doesn’t impact on results as much as we think. Or perhaps the LLM simply plays the law of averages. Co-Pilot of course didn’t know it was involved in a competition - it got the same single prompt every week - whereas the human may have been factoring in games where they could take a risk on a score - and if they got it right would’ve scored more points. If I start thinking about it too much I wonder if even “when” they did their predictions will have had an influence and if they didn’t predict at the same time then the context may have been different (e.g. a player picking up an injury in a midweek game, a manager announcing in a press conference that they’re not likely to play someone). I find it fascinating. I’ll be interesting to see what happens next season and what is the most successful prediction method (and whether Chris Sutton changes anything in his process).

What I’ve been reading

I finished the book I was reading in my last weeknotes and immediately put the sequel on my ‘to be read’ list. I very excitedly then received a notification from my local library that Philip Pullman’s ‘The book of dust volume three: The rose field’ - the final book in the Book of Dust sequence (which is set both before and after the famous ‘Northern Lights trilogy’) was available for me to collect. As I picked up the reservation (it is very much in demand so I’d been waiting around six months) I discovered it was 621 pages (I’m normally a ~300 page book reader) and knew immediately I wouldn’t get that finished in the three week loan period. So I have relinquished that, to come back to it when it is no longer in demand and I can have an extended loan.

Having had a recent weekend in a hired camper van (and living all my best Enid Blyton inspired dreams) I have started my annual re-read of ‘The Secret Island’ by Enid Blyton which I want to get through quickly as in the library yesterday I also picked up my next read which I already know I’m going to enjoy (non-fiction)…what a tease…

Do you collect, use or share data?

We can help you build trust with your customers, clients or citizens

 Read more

Do you want data to be used in your community’s interests?

We can help you organise to ensure that data benefits your community

 Read more