“Government is “teeming” with data and harnessing the power of AI to analyse that data should be a no-brainer”. That’s the view of Steve Webb- and he should know. Speaking with LCP colleagues, he explains that while Government is conscious of AI, it is slow to adopt it
” It will take a long time to get there”, he continues , because of the inertia created by a consultation process that hasn’t changed in years and systems that were designed up to 50 years ago.
The data may be there but often it can only be accessed by speaking to coders, many of whom have passed on or well into retirement.
The drift to stronger regulation
In a hilarious section on the drift away from light-touch regulation, Webb delivers a series of apercus based on the theme that no regulator got fired for over regulating.
8 weeks is a long-time in AI , but it’s taken 8 years to get the legislation in place to combat pension scandals that were emerging when Webb was pensions minister.
The fluid and delicate balance of “self-regulation” applies to Government as well as the private sector, when it comes to AI.
Webb cites the caning TPR got from the Work and Pensions Committee over BHS, which has led to the Regulator demanding powers to jail trustees who allow sponsors to get away with paying dividends prior to paying pension deficit contributions.
Credit where credit’s due
The conversation turns to the success this Government has had in attracting big-tech like Meta to invest in the UK, suggesting that – at least in the policing of social media, this Government has created the light-touch needed for Facebook, Insta and others to operate.
Webb, interestingly, wonders whether the hard-nosed executives who run these companies , have been speaking to Keir Starmer “the next Government”. That’s a rare slip from the cunning ex minister and maybe not a slip at all. I sense a Lib-Lab pact emerging!
Regulating the internet is a tough ask, as Webb points out, talking of setting up “the great firewall of China”.
The problem is made worse by “regulatory capture” where regulators find their exit from the civil service is to the firms being regulated. Steve Webb should know!
If we want consistency and not regulatory arbitrage, we need the same regulation replicated. AI can generate such consistency, but this leads to group thinking including institutional prejudices that are created by replicating the past. Webb cites the example of “firemen” who AI portray as “white men”.
Digital trust and the human input
The source of information we get is interesting. Biden’s recent legislation on “watermarking” attempts to create “trust” in the source of information.
But the Government’s white paper on this subject is already sounding out of date, suggesting that Government can never use conventional consultation to legislate.
Instead Webb suggests that ministers, civil servants and regulators embrace the change and get to use Chat GPT and other tools as part of their BAU.
Webb admits that – to edit an overlong article he had written – Webb used Chat GPT to cut his proof from 1200 to 700 words. It had taken Webb , two hours to cut out 50 words and Chat GPT a minute to do the job properly. He said he felt he’d cheated, but then admitted that as a 57 year old man – life was too short to be fretting about these things.
Better laws and better regulation – the prize is great but the journey is long.
The podcast ends with Webb reminiscing about his time as a politician listening to constituents who weren’t happy with Government using their data to function better.
He concludes that Government cannot impose new ways of working without taking people with them. This means Government has to be trusted and that trust takes time (and good behaviour from those in power).
Government cannot realistically regulate on AI without embracing AI (other than to exclude it). So Government’s role is driven by society and in particular industry’s pace of adoption.
Apparently LCP consultants have GPT enabled on their hardware. So do I. Though, the authenticity of this article is based on my just having listened to the Pod. I strongly suggest you do the same (by the way, this article is 700 words long – coincidentally!).