The hit TV program ‘Humans’ is a great story and also makes a contribution to one of the biggest issues that we will need to address for real, probably sometime in the next century. It is a subject we all relate to: what might happen if scientists and engineers can create a machine that is similar to or even indistinguishable from you and me?
This is an incredibly emotive subject and many mega-stars of science and engineering have made erudite pronouncements talking about existential threats and a hypothetical thing called the singularity (the point at which artificial intelligence might outstrip human intelligence). In all of the debate it is important to understand what the computer software is actually doing and how this results in the things we see. So, for instance, an ‘autonomous car’, simplifying massively, will be equipped with sensors that generate data regarding the position, velocity, colour, temperature (etc) of real physical things. This data is often made up a series of data-dots called a ‘point cloud’ which creates a three dimensional model of the vehicle’s environment. This has similarities with pointillist ‘dot art’ produced by Seurat and Signac in the 1880s (see picture below).
This low-level data is passed through pattern recognition algorithms that identify ‘things’. These algorithms, which are increasingly capable, can now track people, vehicles and the road in real time. It is then possible to write a collision avoidance algorithm that says: try to get to the destination but don’t touch anything apart from the road and, if in doubt, stop. The important point to remember is that immediately we read the words ‘people’, ‘vehicles’ and ‘road’ we make connections with all manner of memories. The amazing thing is that we have a shared understanding, a ‘common-sense’, of what these words mean. Taking another example, when we talk about mown grass we have a common-sense of the colour, the smell, the texture and the sound of a lawnmower. These are deeply textured memories that we all share because our physical bodies (nerves and muscles) work in the same way. Even more amazing is that these physical memories are imbued with emotional data – memories of throwing new mown grass and laughing with friends, or lying in the park gazing up into the sky on a sunny day. If you told me your stories I would empathise and tell you mine. The same is true of ‘people’, ‘vehicles’ and ‘road’. For us they are not just data.
Whilst it is not inconceivable that we will be able to write computer code that will emulate all of this, it is really difficult to imagine how a computer will have the same ‘common-sense’ as a human. After all, we don’t know whether a kitten purring or a puppy wagging its tail is, in human-speak, ‘happy’ but we have decided that this is a useful working assumption.
This concept of ‘common-sense’ is of course essential to any conversation about the right-ness or wrong-ness of an action, the causes and consequences. In society we bundle all of this into ‘regulation’, ‘common-law’ and ‘insurance’ and so much more.
Over the last few years a group, of mostly engineers, have been meeting and discussing these issues and this resulted in the publication of two documents: the Robotics and Autonomous Systems (RAS) UK Strategy and the UK Roadmap document. Both are available to download from https://connect.innovateuk.org/web/ras-sig. They are revealing in what they don’t say as much as what they do.
In combination these documents go beyond the headlines by presenting a pragmatic analysis of the market needs and potential in a number of sectors. The RAS UK Strategy presents a five-part plan that, if implemented, could make the UK a centre for exploring robotics technologies and commercial opportunities. These five components are: grand challenges, skills, clusters, coordination and assets.
Before looking at these it may be useful to explain some of the basic ideas which were distilled into a couple of bite-sized phrases.
‘The arms and legs (and eyes) of Big Data.’
‘The next generation of smarter tools.’
The first phrase highlights the importance of data and our ability to collect vast amounts of the darned stuff and transmit it almost instantaneously around the world. This data exists in a non-human world – it exists in an invisible digital space. So the point about the first soundbite is to recognise that RAS gives the Internet a physical form – the Embodied Internet. It is expected that the data will be collected, connected and used to move physical things and people.
The second phrase suggests that people have always made tools to do more. We are fundamentally lazy explorers. Engineers in particular seem driven to come up with cheaper, faster, safer ways of doing new stuff. This doing ‘more’ is vital because the alternative is doing less which might imply less work and more unemployment. Of course no one knows whether in future RAS will increase employment but taking the long view tools have generally enabled us to do different things, create new jobs and become more productive.
Both soundbites indicate that RAS is part of a continuum of interconnected developments. It is not possible to excise RAS or turn it off. It is happening all over the place and will happen in many forms.
There are perhaps two reasons why the UK should grapple with RAS. The first one is that it will be big and disruptive and we cannot afford to miss out in the way we did with the Internet. The second reason is that ‘us-Brits’ have a reputation for being considered, reasonable people who are cautiously optimistic.
The thought behind the five themes of the RAS strategy recognises that it is all about people: skills, clusters and coordination are essential components of a collective response that involves scientists and engineers, regulators and lawyers, actuaries and insurers, policy makers, investors and the public. Grand challenges is the now widely accepted concept of needing to identify a big issue that galvanises enough people to focus their attention and compete to win a prize and the kudos of presenting a solution. The autonomous car, and everything that goes with it, is such an example.
Assets is the least obvious but perhaps most important part of the RAS UK Strategy. In financial parlance an asset is either tangible or intangible and is generally something which has a value for a period. In the RAS UK strategy we also considered the need to identify both types of assets. Tangible assets are real physical places where RAS technologies can be tested: roads, cities, hospitals, nuclear reactors, airports etc. In order to do this, especially for autonomous systems, there is a need to take intangible aspects into consideration such as value models, safety, regulation (verification and certification) and public opinion.
Let us return to autonomous cars and shed light on the strategy. The strategy is reflected in the investment in the trials that will be occurring in Greenwich, Bristol and Milton Keynes in the next few years. It is also being supported by activities behind the scenes to test the underlying technologies.
More recently we have also come to recognise the need for a place which sits between the closed, private test site and an open, public space like a city or a motorway. The closed space is ideal for testing technology but it is less suited to testing the social science of how people will use and interact with autonomous vehicles. At the other extreme, a city is an enormously complicated environment so, whilst it is easy to collect data it will be difficult to conduct controlled trials that compare the benefits of different solutions.
One potential solution is to find a location which is real but managed. An example being the UK Government owned research site at Culham (see picture below). It is an 80 hectare site with 10km of roads and 2,000 adult employees going about their usual business. It is a manageable subset of real life that creates an environment to explore the issues of how people will use and interact with autonomous cars. The fact that it is government owned also creates the potential for dispensations in regulation to allow long term, 24-7, all weather testing that would be difficult and costly in a public space. Identifying a bounded location in the UK also means that it can become a meeting place for interested parties, including expert scientists, engineers, lawyers, investors and insurers from academia, industry and governments; from all over the world.
Dr Rob Buckingham FREng, Director, UK Atomic Energy Authority, Head of RACE (Remote Applications in Challenging Environments)
