Weeknote S3 E14 24 May 2024 – water under the bridge

It was a week – maybe not so notable?

What was happening this week??

• I started the week feeling refreshed after a sunny camping trip to the Pillars of Hercules in Fife.?
• I had a good chat with Tom Watson of The Good Ship on Monday. He and others are working in practical and inspirational ways to open up data and get more organisations sharing their data in helpful ways.?
• On Wednesday, a useful meeting on the Governance Checkup where we crystallised some important details about how the results page needs to work.?
• A really interesting webinar on the problem of AI ‘model collapse’ from Nigel Shadbolt.?
• Useful catchups on our cyber resilience work and a meeting with comms colleagues about our Big Digital Event on Thursday.?
• On Friday, working on some final bits to our new guide to website projects and a catchup with Nick and Dan about emerging AI experiments.?

At one level this felt like a very ‘steady’ week – to the point where I wondered, ‘why bother weeknoting this?’ But one or two folk referenced my weeknotes recently, so you’re getting it...?

What’s going well?

It was inspiring to chat with Tom and hear about how he is tackling lots of the barriers charities face when thinking about sharing data.?I was really inspired about the way that the work is being shaped from the ground up and meeting charities where they are. That feels very powerful.

The governance checkup chat felt really productive. At one level we were recognising a challenge and more things to fix, but I felt we finished the meeting with a clearer idea about how we need the results experience to work.?

I really felt the benefit of expert input and hard work from Rachel Boyd as we worked on the final bits of the mega website guide together.?(Launching next week!)

What wasn’t so good??

There weren’t many ‘Eureka!’ moments this week – but that’s a wet week in May for you.?

My office day coincided with the big rain so I got a bit of a dousing on the way home. Cheers to the colleague who was asking ‘Where’s this rain then?’ on chat.?

What have I noticed??

The ‘model collapse’ talk was interesting. The short version: as more and more AI content is generated, it will become harder to identify and isolate quality training data. This could lead to a rapid decline in the quality and reliability of AI models.

This is a version of the grey goo problem, and it’s fairly simple concept to grasp (I’ve had it in a slide deck since last October). But it was interesting to think through the profound implications, and how this links back to good data practice. It’s also striking that very eminent computer scientists are actively concerned about it.?

Another way to frame this is to say that ‘the internet today is as original/human-authored as it ever will be’ which is a little scary.?

What have I been reading?

Interesting reflective piece in the WSJ on the challenge of making reliable predictions about tech. Turns out some claims by the tech industry are wildly optimistic!?

The discussion on data provenance and the value of originality got me thinking about my first favourite website on the internet: https://www.artoftravel.net/ I found this site utterly fascinating at the age of 17 as I dreamed about travelling the world. I particularly enjoyed its tips on packing extremely light.?It’s amazing that it’s still on the internet, and that the design is exactly the same as it was back in 1997.?