Comment: Industrial Strategy sees Government return to picking winners
The industrial strategy is a curious mixture, though it does signal a significant change in Government direction, particularly for a Conservative party that has been wedded to the mantras ‘the market knows best’ and ‘we can't pick winners’ for decades.
The strategy has picked some potential winners, wrapping them up in the idea of there being ‘Grand Challenges’ that UK industry should address. This is probably where the strategy is going to wind up in trouble, although the proposed winners probably seemed safe bets.
The four Grand Challenges identified in the report are artificial intelligence (AI), the shift to clean energy, the future of mobility - otherwise known as self-driving cars - and the problems of getting old. These things all fit together fairly neatly. The self-driving car needs AI, is probably going to be electric and will help ferry people around who aren’t really fit to drive anymore. It’s the “neatly defined future” aspect of all this that should probably be ringing alarm bells. Do all these things have to come together for the strategy to work and over what timescale will it happen?
The clean-energy aspect of the strategy seems the safest bet. As a source of energy that will be available until the end of the Earth you can’t really beat the weather. The solutions to a population with a higher proportion of the old may be social more than technological. And, when vehicles drive themselves, how many will we really need? Is this a growing industry overall or a case of managed decline? Tesla, for example, is betting that cars will be shared. And if they are shared we certainly don’t need as many. It would be transport for the public. Maybe we could call it ‘public transport’?
Then there is AI. This looks better than a 50/50 bet. Who doesn’t want their computer to be more intelligent? It is arguably one of the more pleasantly surprising parts of the strategy. But we should be winding down the rhetoric on AI being amazing and everything will be more amazing with more AI. It’s easy to forget that AI is a rough copy of what we colloquially call intelligence. It provides results that could have been generated by an intelligent being. It does not require anything particularly close to actual intelligence. A curve-fitting algorithm can be a perfectly good candidate for machine learning if it deliver results that are close to a much more compute-intensive deep-learning network.
For the moment, the US has gone deep-learning mad. It may turn into another example of faddism. Others do not necessarily have to follow this exact path. Techniques such as Gaussian processes can yield good results with far less data than deep learning, which makes them good candidates for situations where you do not have terabytes of data to feed into the training algorithm. Baidu, Facebook, Google and others have jostled their way into a situation where they have lots of data with which to feed their servers. It is going to be more difficult for others to take them on head-on.
The review by Professor Dame Wendy Hall and Jérôme Pesenti on growing the AI industry in the UK feeds strongly into the existing direction in AI and into the industrial strategy. This is a reasonably good approach for encouraging research into core AI. It has the problem that it may simply lead to the same problem that has afflicted UK-based technology for decades: that the research does not yield a large UK-based industry.
To that end, the strategy is putting more money into PhD research aimed at driving a new crop of AI startups to follow on from those that keep getting bought by the likes of Google. When I heard a minister talk about this last week, my reaction was largely, “This is nice, but is it going to help UK industry as a whole?”
The writers of the strategy have thought more deeply about the problem and have put emphasis behind the more mundane aspect of AI: that it can help make existing business and industries more efficient. To that end, rather than focus entirely on the research end of the spectrum, the strategy does promote some ideas that are meant to make AI more accessible to businesses that are not at the sharp end of R&D. This includes the promotion of masters-level courses and, importantly, the retraining and reskilling of people who might have a computing or background in other areas of engineering to learn the techniques of AI.
The devil is in the detail with plans like this. How will existing technical staff and engineers access those courses? What will it cost them? As a strategy, it is the kind of approach that I thought would probably make a more fundamental difference. Encouraging organisations to talk more about their work in this area to improve their efficiency would be welcome, but that’s more at the level of tactics than strategy.
On top of the Grand Challenges, there are changes such as the small improvement to the R&D tax credit that make some sense, though I doubt it will move the needle much for companies who already do not employ it. Those who already use the system will thank the government. The problem is making it work for a wider range of businesses outside the large corporations and startups who have the accounting advice on-hand. There are plenty of SMEs who have the bigger issue of accessing finance to kickstart productivity-enhancing projects to be able to put the credit into action.
One curious omission from the strategy is the aspect of manufacturing changes that will likely affect many parts of industry. 3D printing suffers from the same problem of AI in being today’s favourite technology. However, additive, stratified manufacturing is critical to a country that has been priced out of the volume-production business. Potentially, it is the most disruptive technology we are faced with because of the way it democratises production. Maybe because it allows workers to own the means of production it’s not yet politically acceptable to the current administration, but industries of all types are going to look rather different as that technology takes hold.