For several years Helen Beetham has been developing an increasingly critical analysis of the development and social impact of Generative AI through her blog and newsletter, Imperfact Offerings. Her latest post, Marking the Government's homework on public sector AI, includes an analysis and critique of the UK government's plans for automated marking and autonomous missiles, and making public data 'safe' by selling it to private 'security' businesses, all in the name of 'sovereignty'. It goes further with an in depth analysis of the data industry and the plans by the UK government for the development of large data centres, in close collaboration with the (mainly) American based AI industry. I am not going to even try to summarise the essay, it covers too much. But I thoroughly recommend that you read it.

But the essay is not just a critique of policy and of the dangers of the Generative AI industry, she also puts forward positive developments that could be taken, if these was a will by policy makers. I particularity liked her discussion of what a digital public infrastructure might look like.

Instead of a spurious ‘sovereignty’, suppose the aim was to procure and secure a digital public infrastructure, composed of material elements (cloud compute, networks, data) and governance arrangements that would prioritise the needs of diverse interest groups. These might be local, regional, sectoral (voluntary, third sector, health, education, research), but the priority would be the participation of those least served and most at risk from the current commercial infrastructures of AI.

Helen draws attention to Muzzacato’s Common Good Framework which "offers five principles for securing public agency in relation to infrastructure:

• transparency and accountability.
• purpose and directionality;
• co-creation and participation;
• collective learning and knowledge-sharing;
• access for all and reward-sharing;"

Rikap et Al's (2024) White Paper on Digital Sovereignty suggests how these might be applied in specific policy proposals for public compute (This, she says is her paraphrase of the original).

1. A democratically accountable, publicly-governed digital stack, comprising digital infrastructure as a service, platforms, a public marketplace without lock-ins, and state procurement preferentially from this marketplace.

2. A research agenda focused on digital developments that could solve collective problems and enhance human capacities, with public research agencies funded to counterbalance the expertise in commercial labs.

3. ‘Ecological internationalism as the basis for national digital sovereignty’, minimising the resources needed and the environmental impact (we could add ‘frugal compute’ or system-wide sustainability as a sixth principle to Muzzacato’s).

4. Dismantle state surveillance and regulate against the ‘misappropriation of collective solutions by specific governments’.

There is much, much more. But go ahead and read the essay, and thinkk about the implicatiosn for education.

About the Image

This image illustrates digital transformation gone wrong, where AI becomes a tool for intensified extraction. Workers operate sewing machines endlessly producing streams of spreadsheets and reports. Instead of liberating labour, AI automation can lock workers into more exhausting cycles of output, without increasing agency or rewards.The image is hand-drawn in ProCreate. This image was selected as a winner in the Digital Dialogues Art Competition, which was run in partnership with the ESRC Centre for Digital Futures at Work Research Centre (Digit) and supported by the UKRI ESRC.