Let's hear it for David Lammy. The lord chancellor is undoubtedly correct in thinking that AI (by which he means machine learning executed in simulated neural networks) will have a transformative effect on the running of the courts. The snag is that the transformation may not be in the direction urgently required.
Quite possibly, it will be in the opposite direction.
We have been here before. Three decades ago, during the euphoria of the first 'dotcom' boom, evangelists predicted that the internet was going to transform the very shape of government. Reports with titles like the 1996 Wired Whitehall (of which I was lead author) led the cheerleading for the replacement of paper forms and files with convenient quick and cheap online transactions.
Among the cyber-punks' most extreme predictions were that private or third-sector 'intermediaries' would take over administrative tasks ranging from the issuing of fishing licences to registering births, marriages and deaths. In this tech Utopia the very machinery of government would wither away, leaving the remaining essential taks to be managed from a server under the prime minister's desk.
E-government, we called it. The idea was half-heartedly picked up by John Major's dying Conservative administration, but embraced with alacrity by the shiny modernisers of Tony Blair's New Labour. Blair set a binding deadline for every single function of government - from paying council tax to obtaining permission for a burial at sea - to be available online by 2005. The revolution was to be overseen by a Cabinet Office 'e-envoy' (a role originally created by Blair's US mentor, President Clinton).
Amazingly, the target was more or less achieved. Public bodies, led by local government, learned to deal with citizens and businesses online. Some services improved: does anyone else remember queuing up at the Post Office with paper insurance and MoT certificates to renew their car tax disc? Of course some processes turned out to be more tricky than envisaged: the most spectacular example must be medical records, the digitisation of which was considered to be simply a procurement problem. Another ghastly can of worms was opened by the attempt to create a secure digital identity system.
Read more
But what e-government did not do was transform the machinery of government itself. Our bureaucratic infrastructure of local authorities, executive agencies and Whitehall departments looks much as it did at the turn of the century. Despite repeated pledges by ministers, the civil service continues to grow. Meanwhile, new technology itself creates temptations for public servants to do more, not less. Ask any self employed person struggling with the new 'making tax digital' requirement to file returns four times a year.
AI's ability to create digital flim-flam will be orders of magnitude greater. As we heard earlier this month, courts around the world are already coping with avalanches of AI-generated legal spam. Our already overloaded legal ombudsman is bracing itself for something similar. And while the master of the rolls' vision of an ecosystem of online dispute resolution services may help fulfill unmet legal needs, the easier it becomes to use, the more traffic it will generate. The architects of the NHS eventually learned that there is not a fixed quantum of ill-health out there to be fixed; the architects of online dispute resolution systems will likewise have to deal with an almost infinite expansion in the human capacity for disputing.
None of this is reason to sneer at the government's plans for AI in public services. The Growth Lab approach to testing systems in a safe-space sandbox is particularly welcome indicator that lessons of past fiascoes have been learned. But as we saw with the adoption of mainframe computers in the 1960s, personal computers in the 1980s and the web in the 2000s the relationship between innovative technology and improved productivity - or falling waiting lists - is always tenuous. True transformation is almost always achieved by new model start-ups, not legacy enterprises, much less public sector ones.
Back in 1996, Bill Gates famously wrote: 'We always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next 10'. Lammy's AI dream looks set to prove that maxim.
No comments yet