Dr Mark Woods is one of our panellists in the afternoon. Mark leads the Autonomy and Robotics group at SCISYS UK. We invited him to tell us more about himself ahead of the conference.
Mark, thank you so much for accepting our invitation to be a panellist. You’re the Head of Autonomy and Robotics at SCISYS UK with over 20 years’ experience in robotics, autonomy, and machine learning. And you’ve developed autonomous/AI/ML based systems for Mars Exploration including the European Space Agency’s first robotic mission to Mars! As a science fiction geek, it sounds to me like you have an awesome job! What are some highlights for you?
Having our intelligent software on ExoMars Rover [named Rosalind Franklin], which is due for launch next year, is a particular highlight. It’s been a hugely productive and collaborative effort between industry, academia, and the European Space Agency. SCISYS will be the first European company to have developed flight software of this kind which provides core capability for the rover whilst operating on the surface of Mars. Whilst it’s been a long journey, we’re very aware that not many get the chance to contribute to a potentially once in a lifetime mission like this so I’m really pleased that we’re approaching this major milestone.
More generally, it’s been the opportunity to work with incredible people to develop and advance technology in our area. As an innovation lead, your focus starting out is on solving real and hard problems for end users and clients by developing first of its kind tech that has impact to move the needle in a noticeable way. For space that has meant making future Mars rovers smarter and autonomous so that scientists can do much more capable exploration work, which is likely to be hugely significant as they search for signs of life on the planet and understand how the Martian climate has changed. In non-space we are using the same technology to keep people out of very dangerous environments, increase commercial efficiency, and contribute to the resilience of our water-bearing infrastructure. Solving that problem and scaling it in a robust commercial way is continually exciting, challenging and satisfying.
Having our intelligent software on ExoMars Rover, which is due for launch next year, is a particular highlight… SCISYS will be the first European company to have developed flight software of this kind which provides core capability for the rover whilst operating on the surface of Mars.
We’ve been lucky enough to have a few technology firsts as a consequence. For example, the first civilian space flight of a Machine Learning system and the longest European fully autonomous traverse of a Mars Rover platform in representative conditions. We’re the first European team to put software of this kind on our first mission to Mars and we’ve now been able to commercialise some of the wider autonomy work we’ve done by leading an inspection of the world’s deepest tunnel of its kind, using a fully autonomous robot and state-of-the-art deep learning technology. I’ve been very fortunate to have had the opportunity to meet or work with some of the brightest and best at what they do over the years: astronauts who walked on the moon or led that programme; a whole host of incredible innovators, engineers and scientists at the European Space Agency, NASA, industry, academia; and most importantly, my team, who continue to solve hard problems in ways that I didn’t think were possible at the outset.
During that time work has taken me to a few interesting places. It will be hard to forget photographing the incredible display of the Milky Way in the remote desert regions of the Atacama Desert. Having dinner under the nose of one of the remaining Space Shuttles in LA won’t be forgotten either!
When we met earlier in the year over coffee, we had such an interesting discussion about AI and algorithmic decision-making in particular. Can you share your perspective on AI with our delegates?
Where to start with this one! To simplify things, I would say that the advances and capabilities offered by the technology have happened more quickly than anyone could have imagined. Consequently, a whole host of questions regarding the potential to change how we work and our rights as individuals, interactions with government and other entities have emerged. There is both huge potential to improve our world but, as with any disruptive technology, there are potential side effects. In some cases, capabilities resulting from tech are arriving faster than our ability to put a framework in place which allows us to manage things in a way that is understood and agreed by society as a whole.
It’s interesting that some of these questions have been debated by philosophers for hundreds of years without clear resolution but now they are a reality and we need to respond to this.
I think it’s interesting that some of these questions have been debated by philosophers for hundreds of years without clear resolution but now they are a reality and we need to respond to this. In some cases, the technology has the potential to allow much deeper and more impactful interactions between our private selves and third parties such as government and businesses as they extract more insight about our behaviour and use this to influence our lives in a variety of ways. The answer lies not so much on focusing on the technology itself, but in consolidating our views on how we wish those public-private interactions to play out regardless of the actual tools that are used and then finding a way to implement an acceptable framework to guide that at a societal level. I think it’s vital that the general public have an active voice in this debate and find a way to take responsibility for important aspects of it. Perhaps this could be discussed at the conference.
It can be used to great benefit for a whole host of challenges that affect us all such as inequality and climate change. The risk of not using the potential the technology offers is greater than ignoring it…we have powerful tools at our disposal so we should be open to actively using them in meaningful ways.
Although there are understandable concerns when it comes to some uses of the technology, I think it is also important to bear in mind that it can be used to great benefit for a whole host of challenges that affect us all such as inequality and climate change. The risk of not using the potential the technology offers is greater than ignoring it, which I heard someone say recently in relation to the UN Strategic Development Goals. We have powerful tools at our disposal so we should be open to actively using them in meaningful ways.
Another issue is the need to understand human preferences and intent when we use algorithms for decision-making in the presence of uncertainty. Getting this wrong can lead to sub-optimal decision making. Stuart Russell, one of the early AI pioneers, makes an important point about uncertainty in the objective and balance in the loss functions used to direct algorithmic behaviour. It is often assumed that the objective (the optimisation goal that an algorithm must try to achieve) and loss functions (which steer algorithmic learning and behaviour) are static and well-balanced.
Stuart Russell, one of the early AI pioneers, makes an important point about uncertainty in the objective and balance in the loss functions used to direct algorithmic behaviour.
In reality, there is often a huge amount of uncertainty in the objective itself, i.e., we don’t fully know what we want until we engage, execute some decisions and process that feedback. Continual interaction is required to refine and define the objective as we make key decisions. Likewise not all misclassifications or losses in algorithmic terms are equal when we train our algorithms even though many of our mainstream approaches treat them as such. Infrequent edge cases can have disproportionate outcomes and this needs to be accommodated. Assuming fixed objectives, when this is not the case and equal consequence during training can adversely skew the quality of the algorithmic performance and lead to poor predictive performance by machine learning algorithms.
There are new research directions being opened up by Stuart and others to address these issues which I think will become more important over time and improve the overall performance of the technology.
There’s a lot of hype around AI. As someone who’s been in the industry for two decades now, what are your thoughts on this ‘fourth revolution’?
I read recently that 40% of AI start-ups actually have no discernible AI in their approach, but it helps them get funding given the popularity of the technology with VC’s (!), so this does add a lot of noise to the debate. At the heart of the hype issue is a bit of paradox.
On the one hand, we have made undeniable progress in areas like image classification, image-based object detection and speech recognition that were not possible a few short years ago. For many years we struggled to develop robust answers to these problems and earlier efforts faltered. Then huge progress was made in a short space of time. This work is certainly not complete and it is worth bearing in mind that recognising things in images is still a narrow task in terms of intelligence.
On the other hand there has been less noticeable progress in tasks requiring more general intelligence. There is a reason why the introduction of concepts such as autonomous self-driving car technology keep slipping to the right: these require this broader type of general intelligence that we, as humans, take for granted. Human drivers are certainly not perfect drivers or control systems, as the accident statistics show. I think we must use technology to improve on the status quo. That said, and aside from our ability to adapt our environmental perception to a range of new environments, we do have an enormous ability to do multi-step look-ahead prediction and planning in complex environments where there is uncertainty. This is a much more challenging task and still an open problem and one reason why delivery has been slower than the hype might have suggested for some high-profile applications.
Behind the higher-level hype, game-changing progress is quietly being made in many lower-profile applications. By automating vision and speech we are able to create an incredibly wide range of applications that impact our day-to-day lives.
However, those narrow tasks of vision-based classification and speech recognition are absolutely central to tasks that inform many of our high-level decision-making processes. Behind the higher-level hype, game-changing progress is quietly being made in many lower-profile applications. By automating vision and speech we are able to create an incredibly wide range of applications that impact our day-to-day lives. So, although we have only automated one narrow element of vision, it is an important one. It has a larger impact than you might expect because it is central to much of our interaction with the world and each other. In those areas the hype (or perhaps attention is a better word) is justified.
I have a vivid memory of your LinkedIn post about taking robots to the beach, likely to the puzzled bemusement of the locals! What are some of the projects you’re working on at the moment, particularly around the uses of AI and machine learning?
This will mean much more reliable and efficient asset management as society and government face challenges brought about by global warming and population growth.
We are working on a range of what we call space and non-space applications. For non-space we are developing commercial robotic and AI technology which can autonomously survey hazardous or extreme environments without the need to place people at risk. We often use the local beach as a handy test ground for short range tests! Our AI technology is able to detect defects in critical items of infrastructure, such as water tunnels, in a matter of minutes with a high degree of accuracy and consistency. This will mean much more reliable and efficient asset management as society and government face challenges brought about by global warming and population growth.
On the space side we are developing AI based technology which will allow our robots to detect important science targets on Mars in an autonomous manner. This could enable scientists to make ground-breaking discoveries about Mars and potentially other planets and moons as well.
Why are you excited to attend the conference and what do you hope to get from attending?
My sense is that anthropologists are ideally placed to address some of the issues raised by our tech so I am really looking forward to hearing their perspectives and discussing a way forward on some of the challenges we face.
I am very excited to be attending! From an innovation perspective we want to build better products that solve scalable commercial problems and have a meaningful/positive impact on society. As I mentioned earlier, human-machine interaction is central to creating tech that disrupts in a positive way. Whilst our expertise lies in the tech side of that interface, we are conscious that our understanding of the human side must always be validated and developed. This is in my view a multi-disciplinary activity and we are eager to learn more.
My sense is that anthropologists are ideally placed to address some of the issues raised by our tech so I am really looking forward to hearing their perspectives and discussing a way forward on some of the challenges we face. On the flipside of that, I think technology could be used to answer a whole host of challenges that anthropologists may not even realise we could help with. I’m also really keen therefore to hear about their challenges and discuss what we could do to help.
In your opinion, why do you think your fellow technologists should attend the conference?
This conference offers technologists an opportunity to hear different perspectives on our work and vice versa…it will help us build better products that are sustainable and impactful in a positive way…I would say register asap!
This one is easy … Solving interesting and hard problems is what makes most engineers, developers and innovators tick. There is huge potential to use our technology in a positive way and address key challenges facing society today. I would argue that rather than fear its use completely we should embrace the potential for new ways of doing things. For that to be sustainable and meaningful it’s important that we delve deeper into the human side of this equation. In my experience, our profession is alive to both the potential and the impact of the technology as members of society and parents etc. but again we need to decide the meta operating framework for use of technology at a higher level.
I think this conference offers technologists an opportunity to hear different perspectives on our work and vice versa, and that can only be a good thing. It will help us build better products that are sustainable and impactful in a positive way. Engineers are always keen to learn anyway so I would say register asap!
Is there anything else you’d like to tell us or say?
I’m just looking forward to it and keep your fingers crossed for the ExoMars launch next year!
We certainly will, Mark! Thanks so much for taking the time to talk to us. I can’t wait for our delegates to meet you on 3 October.
Photo by Thales Alenia Space-Italy of the ExoMars Rover prototype during the 2nd ExoMars Industry Day in Turin, Italy.