Robots are coming – progress or peril?
Robots have fascinated humans throughout history. Back in the 18th century, the “Automaton” attracted huge crowds at exhibitions across Europe as it seemed to outwit human chess players. It was later exposed as a hoax – a person was moving the pieces beneath the table.
Even in modern day, things aren’t always what they seem. In 2021, Elon Musk showed off his new Tesla robot, which turned out to be a man in a robot suit. Subsequent demos of the Tesla Optimus in 2024 turned out to be remotely operated.
But among the flashy faux demos, there have been some incredible breakthroughs. Millions of people have watched videos of Boston Dynamics’ Atlas robot jumping on tables and performing backflips, while a video from Chinese robotics firm, Unitree, showed a humanoid robot demonstrating Kung Fu skills.
Robots are already standard in factories, warehouses, and even hospitals. But what industries will be disrupted next? Will we see them in our homes? And what are the obstacles that we need to clear for us to benefit from, and live in harmony with, these intriguing machines?
In this essay, TFD spoke to Mike Lawton, Co-Founder and Director of Oxford Dynamics, an AI and robotics company, about how he sees the future playing out.
The industry has come a long way in a short time, he told TFD.
“We have been working seriously in robotics for about 70 years, and of course, part of that time has been without advanced electronics. Now, we have some absolutely phenomenal computing power. We have a much better understanding of control loop algorithms (the feedback systems that allow robots to adjust their actions in real time based on sensor input). We have pretty much overcome the mechanics of getting robots to move in exquisitely elegant ways,” he said.
The real challenge over the next decade will be working out “how we get them to behave properly and reliably in a very controlled way”.
This means learning to navigate unstructured environments like homes, which might have objects in unexpected places.
It will also require a programming challenge to make sure that robot brains, powered by AI, really know their limits.
Currently, robots are trained on deterministic models, which produce consistent, predictable outcomes. This is in stark contrast to current generative AI models.
“The Kung Fu robot uses quite a bit of AI, but as we’ve all seen, if you play around with ChatGPT, it isn’t deterministic, it is probabilistic. We can estimate the probability of getting an answer, so there’s always a percentage of uncertainty in what might happen. So when you type a question into ChatGPT, you might not get the answer you were hoping for. Imagine that in robotics. You’ve got a robot standing next to you in a kitchen with a large kitchen knife. If that is running on a probabilistic engine, that robot may, on some occasion, unintentionally cut you with that knife.”
Nobody wants to perpetuate the stereotype of the killer robot. But, if the industry can get it right, robots may help solve one of society’s most pressing problems: how we care for our ageing population.
“Our demographic globally is getting much, much older. In the West, we tend to live more solitary lives than we do in other nations, and the evidence shows we would much prefer to stay in our own homes than go into formal care. There’s a problem that needs to be solved. And I think we will certainly see the emergence of home assistance robotics, starting in a pretty crude way with AI-enabled home assistance for elderly care,” he said.
Unlike the robots in factories, warehouses and hospitals, care robots are likely to take a more human form. Partly because a human-like presence will be more reassuring for an elderly person, but also for practical reasons.
Moody robots?
“It makes a lot of sense to have a humanoid form factor for a robot in a domestic environment. We’ve designed our houses and everything we use in them around the human body. For a machine to be truly useful, I think robots in the domestic environment are probably going to have to have that human form factor, but with greater degrees of freedom, like the ability to swivel 360 degrees at its hip joint to give it greater utility and assistance to a human.”
Getting the robots to safely move around our homes will just be the start of the tick list for people programming these machines. They will also have to grapple with some pretty complex ethical questions, thinks Lawton.
“Imagine this scenario: the robot detects a house fire and can only save one person, its elderly primary carer, or the grandchild who’s visiting, who has her whole life ahead of her. So, who does it choose?”
“It’s an almost impossible problem to try and resolve, but we’re going to have to try and resolve it when we’re putting machines in these situations. And who gets to decide? I don’t know how we solve that problem. Do we look at prevailing law? Do we look at case histories from trials and try to embed that in our software? Our laws and values are different from those in the Middle East, Pakistan, or India. What does that mean for selling internationally – does the software need to be redefined when it goes to different territories? It’s an immensely knotty problem.”
And we probably don’t want our robots to replicate all human behaviour either, as we learned from Marvin, the depressed, pessimistic, cynical android from The Hitchhiker’s Guide to the Galaxy.
“We are the sum of our lived experiences, both good and bad. Do we really want a robot to get temperamental because it’s grumpy? We probably want the better qualities of what we define as human characteristics. But then the question is, who gets to choose? There’s a lot of work we need to do in social science and psychology about picking the properties we want in a robot.”
What is clear is that we need to hardwire into robots something similar to the fictional laws devised by Isaac Asimov. The first of these states that robots may not injure or allow a human to come to harm.
We are the sum of our lived experiences, both good and bad. Do we really want a robot to get temperamental because it’s grumpy? We probably want the better qualities of what we define as human characteristics.
Giving robots a brain
In the next few years, Lawton predicts we will see robots move into what he calls “semi-structured environments”. Unlike highly controlled and predictable factory floors, these include more variability, such as deep-sea exploration and environmental monitoring of forests.
His firm has just developed such a robot called Strider. It has been designed for use by the UK’s Department for Environment, Food, and Rural Affairs in cases of chemical, biological, or radiological attacks.
“Strider can go into a semi-structured environment and map it to quite exquisite detail using Lidar and camera systems. The operator can be a mile away, and Strider can recover a whole range of things from the environment. It can take fluid samples, and it can spray a decontaminant. It can do all this for about nine hours continuously beyond line of sight, under a whole range of conditions,” he explained.
Oxford Dynamics has also developed an AI platform called AVIS (A Very Intelligent System), which aims to bring human-like cognitive understanding of complex data sets and make them easier to interact with. The platform has been used by the British Army, with analysts saying that it can do in 40 seconds what it takes a human eight hours to achieve.
“Ultimately, we want to converge the technologies and put a variant of AVIS inside Strider, so we can have a natural language conversation with a complex machine and achieve predictable results,” he said.
If you really want to go down the robotics path as a start-up entrepreneur, you have got to be absolutely laser-focused on the problem you are trying to solve.
“Everything is about cost and functionality”
This combination of AI and robotics is the long-term goal for many robotic start-ups, and it is an area that is starting to see a lot of interest from investors. But making robots is an expensive and time-consuming business, and we have seen some spectacular failures. For example, Pepper, a cute little humanoid robot that made headlines a few years ago. Pepper was lauded as a robot guide for stations, shopping malls, and events, but it was decommissioned in 2021 due to weak take-up.
Lawton has some advice for start-ups.
“We primarily positioned ourselves as an AI and robotics company, and we tend to lead on the AI aspect. This is because investors much prefer software companies, as the time to develop software is a lot quicker and cheaper than building expensive machines.
“If you really want to go down the robotics path as a start-up entrepreneur, you have got to be absolutely laser-focused on the problem you are trying to solve. Every man and his dog is trying to be the first to produce that human-shaped robot, so forget that – billions of dollars have already been invested by your competitors. What is the niche problem you’re solving? Everything is about cost and functionality.”
Areas he sees that robotics could address in the more immediate future include security and farming.
“Imagine having a robot patrolling an industrial estate with an advanced machine vision system, an ability to take much more natural thought processes and be able to report back in human language what it can see.”
“And in farming, why are we sending a bloke on a quad bike and some dogs to herd cattle when that could potentially be done by an AI-driven robot?” he said.
For some of the more delicate farming jobs, such as fruit picking, there is more work to be done on replicating the complexity of human touch, he added.
“Today, the materials are rigid, and they are not particularly good at working delicately, such as picking soft fruit. If I hand you a plastic coffee cup, the calculations that your brain is doing are actually quite complex: you want to grip that cup, but you don’t want to crush it, so you grip it with just enough force. And as humans, we can walk and talk and drink coffee from a cup.”
“There’s no way that you could hand a robot that hasn’t been exquisitely programmed, a plastic cup of coffee without a disaster on your hands.”
Sex sells
Providing robots with human-like skin and features has already been cracked by those making sex robots. These bots, which have rudimentary movements and some AI capability, are already being sold around the world.
It may seem distasteful to many, but it is a part of the industry that shouldn’t be dismissed, thinks Lawton.
“Sex sells. The biggest driver of the internet was the pornography industry. It drove the need for high-speed video and a whole bunch of other things. And now the sex robot industry is driving things such as tactile skin and solving that uncanny valley challenge,” he said.
“Uncanny valley” refers to the phenomenon where humans are repulsed by seeing versions of themselves in machine form. Making robots look human works up to a point, but cross the line and they quickly become creepy.
A study by Brookings found that just 16 percent of people said they felt comfortable with robots, and there will be much work to do in the next decade to increase public confidence.
For now, we remain fascinated, but wary, of the potential of robots. In future, we may come to rely on them.
Looking for a robotics PR agency? Explore how to build trust in robotics with PR and comms: https://www.wearetfd.com/services/robotics-pr
