The Oxford Robotics Institute investigates systems and applications across domains. Source: Ori
Nick Haws is at the forefront of robotics and artificial intelligence. As Professor of AI and Robotics at Oxford University and Director of the Oxford Robotics Institute, he leads research redefines what robots can do.
With a career spanning indoor service robots, underwater vehicles and robotics in a nuclear environment, Haws brings visionary ideas and basic experiences. He is passionate about the fundamental models, autonomy, and the practical challenges that come with integrating AI into business.
This exclusive interview with the Champion Speaker Agency explores the organization's most transformative technological breakthrough: AI, which is deeply embedded in the workplace. Autonomous robotics are already making an impact there.
Do you consider it the most transformative for today's business from your perspective as a robotics and AI researcher?
Hawk: At the moment there are a lot of very exciting technologies for both artificial intelligence and robotics. In the case of robotics, one of the most exciting things for me is that the autonomy of robotics is approaching business as usual. These are robots that can operate on their own without direct human intervention, using AI to make decisions on board.
Nick Hawes is Director of the Oxford Robotics Institute.
These occur in a very limited range, but are usually used for logistics and so on, but are very common these days, for example, quadrupled robots and drones are automatically flying around the site, looking for changes or issues that may require further testing from humans. From a robotics perspective, such autonomy is very interesting.
Looking further, there is a ton of excitement about humanoids. If I wanted to bring robotics into my business now, I'm not looking at humanoids unless I really want to take risks. However, within the next five to ten years, there may be several use cases for humanoids.
Beyond that, in the wider AI range, there is a great deal of excitement in the basic models (large language models and vision language action models) and effectively compresses all knowledge of the Internet or specialized dataset into something that can be queried very quickly.
Robotic people can use it to understand the scenes around the robot, allowing them to interact better with the world and humans, or simply give the robot the more general ability to act in an otherwise unstructured environment.
Growth of autonomy helps robots reach their potential
I have been working on robot projects in very different environments. Can you share some of the developments that best showcase their potential?
Hawk: Over the years I have deployed autonomous robots in a wide range of different locations. Some of my early works were considering the deployment of autonomous mobile robots. [AMRs] In an indoor setting. The robots were put into the office for security and patrol tasks, and also in nursing facilities and hospitals that support nursing staff.
For months, these robots operated autonomously at once, without the need for humans. They were really autonomous, but they were able to perform only small ranges of tasks. Since then, we have deployed robots.
We operate autonomously underwater robots on Loch Ness and have colleagues here at Oxford and the National Center for Oceanography. The robot collected data from a network of sensors.
They also performed inspection tasks on Sellerafield, including autonomous inspections of robots operating in radioactive environments around the outside of Calham's jet fusion reactors, and Calderhall power plants under decommissioning.
Beyond that, we deployed robots in forests and grasslands. Everything from nursing homes to nuclear reactors made the robots operate autonomously in every region.
https://www.youtube.com/watch?v=hg6mnjg7m6u
We are still learning to use AI
What do you think when AI is integrated into daily workflows, there are important opportunities and risks that organizations should know about?
Hawk: Perhaps the biggest scam is that you are very unsure how to use AI. There is a considerable risk in introducing this into your workflow as you don't really understand some legal aspects, such as copyright.
Honestly, one of my biggest concerns is current energy requirements. Anyone using AI is really contributing to the climate crisis. We all use a lot of electronics, but the energy costs of AI training and reasoning tend to be overlooked by people.
So when you see your carbon footprint as an industry, I want to know how AI is built into it. People are good at dealing with some of the more widely known shortcomings of AI, such as hallucinations and unpredictability. Many people are considering ways to focus on AI, particularly the use of language models in a specific way, and constrain the output to a rationally predictable area.
That's the real advantage – when thinking about chatbots, data retrieval, visual design, code, and document prototyping. Previously, many of these tasks were not impossible to automate, but they were extremely difficult, but the type of AI we are looking at now allows us to automate a wider range of tasks.
For example, query large, unstructured documents that interact with customers on very specific topics. This can be done in a much more general form, allowing you to do a variety of tasks.
Looking back at automation five or ten years ago, these systems were extremely stiff and structured, using chatbots and app scripts. They could only interact in a specific way, and only controlled production in a very specific way.
The advent of these large-scale AI models allows for a wide range of flexibility and generality within tasks, allowing structures with much less input while still being more controlled with output. The approach we are looking at now has real benefits, allowing us to tackle issues that we previously could not address.
But we shouldn't be too engrossed. These are still primarily single shot processes. It may be a single dialogue with multiple steps or a single image generation, but there are not many systems that can autonomously complete a series of individual tasks to achieve the goal.
For example, to book holidays or arrange delivery, multiple independent parts must be coordinated. This is one of the areas where AI systems are currently lacking. This is the ability to plan and coordinate across multiple domains.
https://www.youtube.com/watch? v = evfqbryilca
What core messages do you want us to leave behind about robotics and AI when addressing viewers?
Hawk: “When talking about robotics and AI, and I hope you have that feeling in my other answers – I'm trying to leave it grounded. I think it's important to split artificial intelligence and autonomous robotics. These are important and inspiring tools to use in the future, but we shouldn't be engrossed in hype.
Even irrelevant abilities and identities should not be over-approved. These are software and hardware tools, and you shouldn't suddenly think that they are the solution to everything. These technologies have many limitations.
For me, it's both excitement and ability: telling you both what they can and what they can't, and what you should be careful about. I want people to get a better, more realistic understanding of these exciting technologies and the future we are going with them and get away from my talk. ”
About the author
Tabish Ali is a celebrity content and outreach executive at Champions Speakers Agency, the leading European keynote speakers office. In this role, he leads an exclusive interview campaign with world-renowned experts in AI, cybersecurity, digital transformation, sustainability and leadership.
Ali conducted more than 200 interviews featured at outlets such as MSN. Benzinga, Scots, Edimba Livening News, and Express & Star. His work transforms complex insights from industry leaders, including FTSE 100 advisors, bestselling authors and former government officials, into captivating thought leadership.
