I attended a seminar the other day which raised some questions in my mind about the next and prior waves of automation, the location of value creation and the legal/social barriers to adoption. Much is spoken of the use of artificial intelligence to augment or replace professional workers and this note briefly looks at this. It examines the nature of decisions and the need to transparently serve a human rights agenda, the question of regulation and assessment by one’s peers, and why it’s so hard to organise Trade Unions amongst the software authors.
Lawyer’s first! I have been arguing that the states apparatus for making judgement must be transparent and as must its software be. This is recognised in the GDPR and by its predecessors via its “freedom from automated processing”; wherever decisions are taken by computers they must allow for human review. This is true in the private and public sector. The consequences of getting it wrong are often critical to the individuals concerned. In court proceedings, natural justice requires that the evidence is assessed by a jury of one’s peers and that any decision by a court or judge can be appealed. Code becomes evidence and replaces juries of peers, even open source code requires expertise to assess. Humans cannot be replaced by code in these processes and we need to remember the concept of a jury of peers. AI cannot be cross examined in a court. Justice requires humans who can see the provenance, linage and evaluate the chain of custody of the facts & evidence. The whole of the first episode of Blake’s 7 is about how the totalitarian Terran Federation rigged a computer based trial in a legal system that had sapped the will of the legal profession to even provide a defence.
One would have thought that medicine is an absolute no go for the use of AI, but at the edges doctor’s and nurse’s professionalism is being chipped away, for instance by DWP when performing “fit for work assessments”; doctors have been written out of the process. Part of this is the diminishing of the meaning of the word professional; when looking for a picture to decorate this article, I typed professional into unsplash’s search engine and got pictures of a bunch of office workers. Michael Gove’s “we have had enough of experts” may be one of the most destructive political comments ever made. We forget that one inhibitor to replacing professionals with software is that to act as a professional requires that people are tested, certified and subject to regulation by their professional peers. I.T. and software engineering are not professions despite the efforts of the BCS & IEEE. In order for A.I. to challenge people in this professional work, we’ll need to reconsider our approach to certification and regulation of software engineers and software itself.
Why is it so hard to organise I.T. workers? Part of it is their exceptionalism; part of it must be the 21st century version of the “Labour Aristocracy”, compounded by the outsourcing, abolition of work places and the ideological victory of anti-solidarity in the workplace epitomised by stack ranking performance management systems. We should acknowledge that the mergers that have taken place amongst the Unions will have accelerated the bureaucratisation of the Unions and diluted their brand as experts in an industry. The domination of US companies with anti-union managements can’t help, but for some reason even when the datenkraken become organised activists start limited staff associations. The surviving manufactures such as Fujitsu still have recognition agreements and the drive to outsourcing public sector business has led to recognition agreements being inherited via TUPE for the strangest of companies such as HP, the scion of Ross Perot’s system integrator business. My confusion about the fundamental economics of software may be because it is not well understood. While some of my earlier writings suggested that software was only constrained by imagination, I have changed my mind as it is clear that the speed of light is a factor in response times and all computers are made of natural resources and require electricity to run. I am still reading and thinking about this and my notes are in my wiki article, the Fragment on Machines.
The need for electricity reminds me of Carlota Perez’s observation that the coming of the railways Increased the demand foe horse drawn miles in order to get goods and people to the stations; the demand for I.T. needs electricity which sustains a demand for oil (despite its current zero price), the innovation technology driver of the previous long wave. Even the existence of deliveroo who use bicycle couriers to deliver food, shows the continuing longevity of the need for delivery and its retro technology, although let’s wait for drones to become cheap enough.