Weeknotes

Weeknote #35: Sandbox

Portobello beach.

What happened this week?

A slightly shorter note this time as I emerge from a recruitment-heavy week. We interviewed for a couple of roles linked to a UKRI-funded project and I then sat on the panel for a Director position, alongside Amina Shah and Dr Jess Gardner from University of Cambridge. The long days were helped by the incredibly high standard of candidates and noting how experienced interviewers can tease out the right information through simple, well-timed questions.

Later in the week I had a call with Chris Michaels and the N-RICH team on using the Library's open datasets across their prototypes. There's recognition of the overlaps with other funded programmes and they're asking all the right questions to keep the focus on developing national research infrastructure.

Thinking about digital infrastructure, the CSPS website continues to be a shambles. I'm usually very forgiving around big online service launches but even the basics of a load-balanced high-use site aren't in place. One comment summed it up: ‘Everybody agrees it’s appalling, but there’s no end in sight’. (Also, updates about the service are hosted on the website that is currently not working.)

On Thursday I had a super interesting chat with the marketing team on how we apply our AI Statement to the procurement of creative outputs - when should we allow the use generative AI, and where are the red lines? Staff awareness is growing and more questions are being raised about technology choices and their impact. I'm keen to avoid a proliferation of statements created with good intention but leading to possible unintended consequences.

The week ended well with a Friday morning coffee meet up with fellow digital men, John Fitzgerald and Hugh Wallace. We pondered the increasing tech burden, step-count conversions, conspiracy theories and a possible AI event for culture- and charity-sector folk. I'm also keen to see John's work-in-progress guidance around the environmental impact of AI.

Interesting things

The Guardian jobs site is closing, as noted by Sam Scott-Wood who points to Culture Briefs for sourcing opportunities in the cultural and heritage sector.
The Scottish Government has published new Artificial Intelligence (AI) guidance for the public sector which, at present, is a list of well-considered questions to ask before any AI engagement.
The agents are coming and with it an accountability gap. An interesting post on the need to record what AI agents do, rather than taking their word for it.
Jude McCorry from Cyber & Fraud Centre Scotland on why the fundamentals still matter in the face of an AI-driven cyber threat.
The Harvard-based Institutional Data Initiative have published a dataset of 1.47M newspaper pages and a new processing pipeline in collaboration with Boston Public Library.
The Library of Congress has blogged about its AI Sandbox for staff which enables small-scale experiments in areas where tools can "exceed the information processing capabilities of even the most energetic human". This approach echoes our own with locally-hosted small language models and we've had chats with the Rijksmuseum and others looking to establish "test and learn" environments. Key for me will be organisations sharing outputs openly for others to build on.
The Australian Museum have launched AI-assisted search focussed on visiting the museum. It's interesting to see them label this as "AI search" upfront on the homepage and give the tool a human framing ("I can answer questions...").
This pleased me but I'm not sure why: Mom bought 150 lemons.