Unpicking the Artificial Intelligence Talent Arms Race
In the Winter 2017 issue of Breakthrough, UKSPA Chairman Dr David Hardman MBE highlighted the technology skills gap and said that ‘companies requiring specific skills will either struggle to grow, or will move elsewhere to access the necessary talent. This article explores how AI is driving the skills gap, and what can be done about it.
The rapid adoption of AI is accelerating the broader digital skills gap, and with one new AI company being formed in the UK each day there is now an ‘arms race’ for talent. At a recent AI conference in London[i], almost every presentation featured skills development as an imperative. This is in line with a recent CBI report which stated that the biggest barrier to AI adoption in UK companies is the shortage of AI skills, with two thirds of pioneering businesses saying they don’t have the skills and capabilities needed to adopt AI.[ii]
But what skills are required, how might they be developed, and what specific actions should businesses take?
AI IS THE NEW JAVA
AI is the application of mathematics to data. The AI process involves inputting data, processing and optimising it, applying mathematical models, and obtaining a prediction. According to Paul Clarke, CTO at Ocado, “AI is the new Java, it will be everywhere and in everything”.
AI differs significantly from other technologies in that it can develop itself without human intervention. As we race towards the singularity – the point at which machines exceed human intelligence – we can expect increasing ‘self-determination’ in intelligent systems. This means that at the sharp end of AI development people are required to design systems that program themselves, so the level of abstract thinking, skills and knowledge needed is PhD level – even if AI developers don’t have a PhD, their thinking needs to be broad and deep, and they need to be capable of significant abstraction.
AI is not new technology, and many AI processes are now automated – e.g. AutoML allows the automated selection of the mathematical models that are applied to the data, and other tools can be used to automate workflows and other processes like feature selection in the data.
However, automation in AI processes can only go so far, and between 60% and 80% of the AI process is the preparation of data. So, skills like data wrangling, cleaning, model validation and data visualization will continue to be in demand.
AI SKILLS CONTINUUM
The core skills needed to build and work with AI systems are mathematics (particularly statistics and probability), data manipulation, and computer science. Depending on the focus of the business, deep domain expertise can add a lot of value to the AI development process.
Governments in the West are increasingly demanding more openness, and this means showing your workings – including algorithms – and this will drive demand for strong communication skills.
Is A.I. becoming the new Latin – open only to an extremely skilled elite?
Only a very small number of people will be required to develop AI solutions. As AI reaches into the workplace, most people will be either managing the use of AI to produce value, or they will use AI to augment their work. For example, one can imagine how staff in a company could talk to an AI assistant to help make decisions, but a comparatevly small team of people would actually implement and support the system.
There’s also a real need to educate everyone to use AI in a smart way. This means people generally being savvy enough to not be persuaded by AI assisted fake news, or aggressive marketing, and being able to analyze how algorithmic decision making affects them. Without a basic understanding of how AI works, there’s a risk that many people will simply have AI done to them.
Higher Education, training and self-learning all have roles to play in ensuring that A.I. is used in the workplace. The UK Government is active at the top end of the AI skills continuum. For example, government funded Alan Turning Institute has a fellowship program designed to attract the brightest and best in the field. The EPSRC are funding clusters of PhD level research in AI.[iii] Other initiatives include funding for 1-year conversion Master’s degrees.
At the lower end of the workplace AI skills continuum Further Education and workplace training can have significant business impact.
Many people see A.I. as a no-go zone. This has to change.
A key skill required in all businesses is to recognise where AI can potentially add value. To take advantage of AI, there must be people in the organisation who know what AI can do, and understand what products and services are required. There must also be people who can visualise AI use-case scenarios, gain support, obtain resources, implement AI solutions, and manage change.
AI WILL BE COMMODIFIED
As with most technologies, it’s a reasonable expectation that AI will become commodified, and that general-purpose AI tools for business will become widely used. In 1962 the concept of a spreadsheet was embodied in Fortran[iv] – a language accessible only to specialists. By 1985, the spreadsheet had become fully visual in the form of Excel for the Macintosh. Now Excel is ubiquitous. It’s not unreasonable to expect a similar journey with AI.
Peering further into the future, we evisage Bio-Inspired AI fundamentally changing traditional development paradigms towards Goal Directed Design, or GDD. This will enable developers to tell the computer what is needed instead of what to do, and this has implications for the kinds of skills that will be needed in all kinds of software development.
WHAT SHOULD BUSINESSES DO ABOUT THIS?
AI can be used in an exceptionally wide range of scenarios, from swarms of microscopic robots to big data processing in massive server farms. So, understanding the scope, scale and range of benefits that AI can bring to business is an essential first step. The most important question to start with is “what problems can AI solve – for our customers, and for the company”.
The next logical step is to appoint a team leader to drive an AI transformation program, starting with the ‘lowest hanging fruit’, and then driving a cross-company program of AI awareness and skills. At the same time, its important to ensure that the data upon which AI services are built is of the highest possible quality.
Of particular importance are leadership skills, and leaders should know not just about AI itself but the kind of technologies that can compliment or underpin AI systems – platforms & architectures; IoT; Blockchain; big data; and analytics.
To take full advantage of AI, companies also need to develop skills in designing digitally-based solutions; solution selling; effective cross-team collaboration and innovation process.
Finally, its worth realising that AI is just a tool – albeit a very powerful tool, of course. Whilst it helps to be able to think at PhD level to undersand how to develop AI systems, for most companies the heavy lifting involved in implementing AI will be about definining the problems and opportunities, choosing and implementing the right solution components, and helping people incorporate the power of AI into their work.
[i] Westminster eForum, AI and robotics, innovation, funding and policy priorities, February 27th 2018
[ii] CBI, Adopting the Future survey, 2016