Not my commonplace — presentation at GR’24 — Find out how to Crack a Nut – Cyber Information

I’m attending International Revolution 24 on the College of Nottingham. That is at all times an attention-grabbing convention and a very good platform to meet up with colleagues and mates from all around the world. I spent Day 1 attending the 4 workshops of the AI and procurement stream that Jessica Tillipman (GW Regulation) and I co-organised. This was a day filled with insightful and interesting discussions, and I took lots of concepts for additional analysis and coverage work, together with on the necessity for a special method to conflicts of curiosity that may be embedded within the tech (extra on that another time).

My presentation featured a unicorn metaphor to synthesise my views that the present method to the regulation of public sector AI adoption by means of contract and, particularly, by means of the EU AI Act and the pilot mannequin AI contractual clauses is sure to fail and requires a rethink. Like ‘haunting for a unicorn’, pursuing ‘reliable AI’ is a significant problem as a result of we’re after one thing that we now have by no means seen and might solely describe on the premise of our creativeness. The absence of clear metrics to precise open-ended ideas, corresponding to equity, and the impossibility of designing efficient contract-based governance for the socio-technical embedding of the AI, inevitably create extreme indefinition within the regulatory method and go away it open to business dedication and regulatory tunnelling.

Furthermore, the ‘regulation by contract’ technique based mostly on a skeletal blueprint (the mannequin AI clauses) that requires vital quantities of technical and business experience for public patrons to fill in is like asking a baby to repeat a top level view of a unicorn and color it in. The end result could also be endearing and even artsy. However it might hardly resemble a unicorn, which we could solely see by stretching our creativeness. And deploying ‘AI on AI’ won’t resolve the scenario as a result of ‘technical fixes’ can not make up for the indeterminacy of the regulatory information and targets—very like an automatic description of the unicorns drawn by youngsters (beneath) solely approximated the goal by figuring out on image as that of a horse, and the opposite as that of a kid!

Leave a Comment