UK AI report - academic, disconnected, disappointing

The following quote is the opening salvo of a new government report, ‘Growing the Artificial Intelligence Industry in the UK’, produced by Prof. Dame Wendy Hall, Regius Professor of Computer Science at Southampton University, and former IBM Watson VP Jerome Pesenti, now CEO of AI vendor BenevolentTech.

We are at the threshold of an era when much of our productivity and prosperity will be derived from the systems and machines we create. We are accustomed now to technology developing fast, but that pace will increase and artificial intelligence (AI) will drive much of that acceleration.

The document was heralded by a press release announcing that it would unveil plans to “supercharge” the UK’s AI industry, enabling it to become the “clear world leader in the development of AI to boost productivity, advance healthcare, improve services for customers, and unlock £630 billion for the UK economy”.

Ministers lined up to support the report, while Professor Hall added:

Now is the time for us all – scientists, researchers, entrepreneurs, and the government – to come together and address the issues about how AI is going to impact society and seek ways to ensure that we’re able to deliver the great breakthroughs the technology has the potential to deliver.

Bold claims and big words, but does the reality match the hype? Not by a long way.

Whitehall has been talking up AI and robotics since identifying them among the ‘eight great technologies’ that will be critical to future prosperity, but it has yet to back its rhetoric with anything like world-class investment. At the outset, the report appears to offer some blunt advice:

We have a choice. The UK could stay among the world leaders in AI in the future, or allow other countries to dominate. We start from a good position in many respects, but other leading countries are devoting significant resources to growing and deploying AI.

The UK will need to act in key areas and to sustain action over a long period and across industry sectors, to retain its world-leading status, and to grow our AI capability as well as deploying it much more widely.

Or as the recent RSA report, The Age of Automation, put it, the UK is good at moaning about the societal impact of AI and robotics, but bad at implementing the technologies themselves, and as a result may end up losing out in the biggest new market this century.

The report sets out in detail the UK’s prominent place in the history of AI, its expertise, and the technology’s potential to transform sectors such as healthcare, finance, and transport – particularly in London and the South East. But what are its “supercharging” recommendations?

Data Trusts

Beyond summarising the findings of a few recent studies, most notably by PwC – whose expensive, measured, and reassuring tone always pleases ministers’ ears, – this is where a real sense of anticlimax sets in, together with evidence of political naivety.

First, Hall and Pesenti urge widespread improvements in data access for all, in order to make the UK a truly data-based economy. The government should develop Data Trusts, they suggest, to ensure that data exchanges are both secure and mutually beneficial. A good recommendation, but this government’s record on data trust is patchy at best, and lamentable at worst, as the Prime Minister’s surveillance programme and the Home Secretary’s recent comments on encryption demonstrate.

That said, the UK needs to rebuild its reputation in data trust in the wake of these and other misguided assaults, and so establishing Data Trusts may be one means of doing so – assuming that the government’s desire for political expediency doesn’t scupper that good faith on day one.

Hall and Pesenti add that the government should improve the availability of data for developing AI systems:

Government should ensure that public funding for research explicitly ensures publication of underlying data in machine-readable formats with clear rights information, and [that it is] open wherever possible.

To support text and data mining as a standard and essential tool for research, the UK should move towards establishing by default that, for published research, the right to read is also the right to mine data, where that does not result in products that substitute for the original works.

Should these two noble and well-intentioned recommendations become government policy – which seems unlikely – some organisations may respond by burying their research data and only publishing the headline findings.

Skills

Next, Hall and Pesenti move on to skills. As previously reported (see diginomica, passim), as automation and AI increase, low-skilled jobs will be swept out of industries, while smaller numbers of new, skilled roles will be created (10 jobs out, 2.5 skilled jobs in). Without the right skills base in the UK, economic disparity will increase and organisations will find it tough to recruit workers without looking overseas.

The report says:

Government, industry and academia must embrace the value and importance of a diverse workforce for AI, and should work together to break down stereotypes and broaden participation.

That’s an excellent point, with some estimates saying that only 17 per cent of people in STEM careers are women, for example. But then the report gets more specific and – arguably – more problematic, with a limited focus on high-end research:

Industry should sponsor a major programme of students to pursue Masters level courses in AI, with an initial cohort of 300 students.

Government and universities should create, at a minimum, an additional 200 PhD places dedicated to AI at leading universities. As the UK trains and attracts additional academic talent, this number should grow continually year on year.

Universities should explore with employers and students the potential demand for one-year conversion Masters degrees in AI for graduates in subjects other than computing and data science.

Universities should encourage the development of advanced credit-bearing AI MOOCs and online Continuing Professional Development courses leading to MScs for people with STEM qualifications to gain more specialist knowledge.

These are excellent recommendations in themselves, but by explicitly identifying AI as a high-end academic discipline – one reserved for small numbers of Masters and PhD students – the report ignores the much bigger challenges facing the UK: a severe lack of skills at secondary and graduate levels, and the skills gaps already opening up across a host of industries, including manufacturing.

Hall and Pesenti add:

An International fellowship programme for AI in the UK should be created in partnership with the Alan Turing Institute: the Turing AI Fellowships. This should be supported by a targeted fund for identifying and recruiting the best talent, and by ensuring that the UK is open to any and all of the eligible experts from around the world.

Another good idea. Turing’s reputation – which is finally getting the widespread acknowledgement it deserves – should be placed front and centre of UK strategy, suggest Hall and Pesenti. The Alan Turing Institute should become “the national institute for artificial intelligence and data science”, they continue, “becoming truly national and expanded beyond the current five universities, with a key stated aim that centres its mission on artificial intelligence.

The Alan Turing Institute, the Engineering and Physical Sciences Research Council (EPSRC), the Science and Technology Facilities Council (STFC) and the Joint Information Systems Committee (JISC) should work together to coordinate demand for computing capacity for AI research, and negotiate for the UK research community.

The Information Commissioner’s Office and the Alan Turing Institute should develop a framework for explaining processes, services and decisions delivered by AI, to improve transparency and accountability.

Further recommendations

The critical importance of the Alan Turing Institute aside, how else should the UK encourage and support the uptake of AI? The report recommends that the government should:

• work with industry and experts to establish a UK AI Council to coordinate and grow AI in the UK
• order the Department for International Trade to expand its support programme for AI businesses
• ensure that challenges addressed by the Industrial Strategy Challenge Fund (ISCF) and Small Business Research Initiative (SBRI) are designed to attract and support AI applications, and:
• draw on the expertise of the Government Digital Service (GDS), the Data Science Partnership, and experts working with data in other departments, to prepare the public sector for AI and spread best practice.

Again, these are good ideas, but the last recommendation overlooks the fact that the GDS is a neutered internal resource, marginalised by the interests of the very tech suppliers it was set up to counteract. In slow, bureaucratic Whitehall, new ideas need sustained political support to survive.

Finally, the report says that techUK should work with the Royal Academy of Engineering, the Digital Catapult, and key players in industry, to develop practical guidance on the opportunities for AI right across the UK economy. So what does techUK itself make of the report? Head of AI Sue Daley went with supporting the tone of the document, while ignoring much of the detail – a sign, perhaps, that it isn’t impressed.

The independent review commissioned by Government is an important step in identifying where the UK must strengthen its technical, academic, and research leadership to make it the best nation in the world for AI companies to thrive.

If the UK is to realise the full economic and social benefits of AI, the Government must foster greater demand for, and adoption within, both the public and private sectors, and across industries. techUK stands ready to support the Government in this aim.

My take

Aside from its political naivety, this is a well-timed report that makes solid recommendations and doesn’t pull its punches: all good things, even if there is little acknowledgement of the risks of AI. Fortunately, these have been set out in recent RSA, UK-RAS, and techUK studies, among others.

Yet as a summary of the challenges facing UK AI, robotics, and more, it falls well short of the mark. Its proposed solutions, while excellent, are partial and create the impression that the underlying purpose of this report was to make academia’s voice heard in Whitehall and secure new long-term funding.

No one doubts the stature of the UK’s universities and research institutes, or the world-leading AI and robotics expertise within them, but by appearing to regard AI solely as an academic discipline, the report misses nearly all of the most important challenges facing the UK.

These are:

• The country is investing too little in AI and robotics: the Japanese government, for example, is investing between 500 and 800 times as much money over the next five years. While the report acknowledges that investment needs to rise, it does so with little urgency, beyond its strong opening statement.
• 80 per cent of the UK’s funding in these sectors comes directly from the EU – according to Parliament’s own Science and Technology Select Committee. Post-Brexit, that money must be in doubt.
• The UK’s uptake of AI and robotics is falling behind other countries in the EU, Scandinavia, Asia, and North America.
• The current UK government sees AI and robotics solely in terms of improving productivity, and appears to have no interest in growing new markets. The UK needs to be much more ambitious and entrepreneurial.
• The government’s attitude to data trust and security revels in its own ignorance, if comments by the Home Secretary and other ministers are anything to go by.
• Brexit is putting pan-European research at risk, creating uncertainty for all UK technology and research programmes.
• The Brexit iceberg aside, Whitehall is busy arranging the deckchairs on the good ship UK, with no idea of its passengers’ final destination. Even a picture postcard of what the future looks like would help.
• According to multiple reports, the UK is investing too little in secondary and further education, and nowhere near enough in its core infrastructure to compete globally.

These are the real problems for which the UK needs urgent solutions when it comes to AI and robotics. This report does little to address them.