What I’ve been doing
I had a really lovely Easter break (I took a couple of extra days ahead of the long weekend) and ate lots of chocolate. The week after I was the only team member working most the week so I had to resist the urge to be at my desk re-enacting Bridget Jones…
It has been a productive couple of weeks since Easter. I completed the ‘project stand up’ for phase two of our Mohn Westlake Foundation funded work including updating the website. Also updated the website with a PDF version (in English and Welsh) of the Worker voice in public sector procurement of digital and AI systems in Wales toolkit we’ve co-created with TUC Cymru. A learning point here is that despite the limitations of PDF as a format, for some with high security levels on their computers (e.g. local authorities) a PDF is needed.
I’ve done two lots of Data and AI Civil Society Network notes (this week’s are pending review by ICO who attended as guest speakers); published community of practice notes to the members; attended a meeting with the EY Foundation who we’re hoping to work with, and followed that up with an initial quote.
I’ve picked up some logistics management for a couple of events we’re doing in partnership over the next two months. I’ve spent a week getting myself up to speed and getting them as progressed as much as possible ahead of check ins next week.
As I mentioned in the last weeknotes I’m trying to do some more writing. I’ve written and submitted a low risk bid after Easter (unlikely we’ll get it, but not predominantly due to my bid writing) and re-drafted our PVAI Op-Ed (ghost writing, which was a new and tricky experience). I’ve also been working on the draft annual report 2024-25 and strategic roadmap 2025-26 to get them to a point for Jeni (and co) final review and tweaking over the next week.
And I cleared some more admin level tasks including circulating a draft toolkit Jeni has been working on with a group for feedback; adding a speaking engagement for Jeni to the website and scheduling associated social media promotion of that; invoiced for a late PAIRS donation (and updated the accountant re end of year accounting of that).
What I need to take care of
Within the strategic roadmap for this year a goal I am leading on is “All our reusable and externally visible work has been archived somewhere for the long term”. How we approach this is something I’m starting to think about - if you have any experience of doing similar I’d love to hear from you.
What I’ve been inspired or challenged or moved by
There’s a lot of talk about what Big Tech and the Government think technology should do for us. I fully support this request for action.
I also kind of inspired myself in the last couple of weeks. After my last Weeknotes where I shared my thoughts on an article that I’d read, I’ve done a few really short posts on LinkedIn recently sharing my thoughts - or links to articles - about some recent mainstream AI news items. I’ve copied those below so they’re captured here.
- You’ve seen the AI generated images of people as dolls/action figures.
You’ve probably also seen the pushback of human generated alternatives.
There’s some high horses being got on around this trend (and similar others) and I’m not going to do that - but I do feel we all ought to inform ourselves a bit better before engaging with things like this. There’s no such thing as a free lunch.
In my humble opinion this is a good, easy read, article covering the main points.
- “Manners cost nothing”.
Turns out that’s no longer strictly true.
Of course the real message here is that if even just “please” and “thank you” are burning energy/having a significant carbon impact, then maybe think twice before just playing on AI like ChatGPT and others.
- There is nothing “intelligent” about Artificial Intelligence (at least not the publicly available stuff) so we need to keep bring our intelligence to the party.
There’s been two reported examples in the last week of “AI” completely making things up (this is just the selected example LLM, don’t be thinking this is unique to that single product).
🔗 Man told (falsely!) by ChatGPT that he had killed his children (*trigger warning child death)
🔗 Google AI search results explain made up idioms
Large Language Models (LLMs) work by searching everything they “know” (a lot, there’s a lot of data being fed into them) and finding you the most probable acceptable answer. It will rarely tell you it doesn’t know the answer (and usually this is only when parameters have been set for it to not answer questions e.g. in China you can’t get answers to questions about the government) and it may even extrapolate from other facts to answer your question.
There’s an obvious problem with these false replies in a world that is gathering misinformation momentum, when people want to debate facts - there’s now another layer of non-facts being presented as facts. This presents a lot of risks - some minor and some pretty significant.
As a child of the 80’s and learning the internet in the 90’s as it exploded into public usage, I learned the importance of understanding the different quality of websites and applying critical thinking to what I was reading. Heck, I had my own website in my mid teens so I knew that anyone could write whatever they wanted on the internet. The teens in my life just accept whatever the top result Google provides them when they “search it up”.
There is some interesting work being done around using AI to support searching data but in the meantime, I for one will not be using the AI search results provided or ask a LLM a generic question (that I would have used a search engine for). Depending on how important accuracy is to me I will also try and verify the answer from another (trusted) source.
What I’ve been reading
I’m a Big Bang Theory fan. I’m currently reading (and very much enjoying) ‘The Big Bang Theory: The Definitive, Inside Story of the Epic Hit Series’ by Jessica Radloff. It’s probably less enjoyable for my wife who has to put up with my “oh oh oh, let me tell you this” moments!