The robotics market has always been complex, shifting, and uncertain, often characterised by bold public announcements that carry little weight, and less PR-friendly innovations that transform industries and create real economic advantage.
This year is no exception.
Tesla is set to unveil what is now known as the Optimus humanoid, formerly ‘Teslabot’, at an AI event at the end of September.
Often, humanoid robot unveilings are heavily stage managed, with the devices going through pre-programmed, scripted routines to create the illusion of advanced functionality.
At this point, it is unknown whether the event will reveal a physical design mock-up, perhaps with the intention of exciting investors and lifting Tesla stock. Or whether the company has built a working Optimus prototype, after what would appear to be an unfeasibly short development time – at least since the concept was revealed a year ago.
No robotics expert I have spoken to since August 2021 is aware of Tesla acquiring robotics companies, beyond investing in its own automated factory processes.
From the information currently available, the Optimus robot will share onboard technology with the Tesla car line, including a system developed from Autopilot.
CEO Elon Musk – currently engaged in a legal battle over his on/off Twitter acquisition – claims that development of a robot line will take priority over other products and will, ultimately, create a market more valuable than electric vehicles (EVs).
But will it? It’s a dangerous assumption, when the utility of an EV and/or a connected, autonomous vehicle is far more obvious.
Over the years, countless advanced humanoid robots, such as Honda’s ASIMO, have been impressive engineering achievements whose real purpose has been to act as brand ambassadors and showcases of tech excellence.
Optimus’ role as a brand superstar that communicates ‘future, intelligence, and quality engineering’ – and perhaps even ‘hubris’ – would make sense, particularly as Musk himself is more than capable of tainting his own brand.
But beyond the worlds of corporate messaging and engineering/design showcases, it has long been unclear what the real purpose of humanoid devices is, once they are taken out of the realms of science fiction.
Often, they are portrayed as having helper/companion applications, but that will demand real intelligence, security, safety, and reliability for robots to resemble their sci-fi counterparts in anything more than sleek, futuristic designs.
Just as important, it demands that such robots are affordable, rather than hugely expensive luxuries that will depreciate more quickly than a car.
An on-demand, service-based market is likely to develop, but only if humanoids provide services that people need and are prepared to pay for. As a robotics CEO once told me, “Nobody makes money from selling robots. You make money from selling the services they provide.”
At which point the question arises: why not just employ a human?
This should be a warning to Musk: unless your robot can provide viable, safe, essential services to its future owners or managers, Optimus is likely to be a money sink and an expensive plaything. But roll it out once a year to give your share price a kick, or feature it in expensive TV ads for cars and space ships, and you’re probably onto a winner.
Countless humanoid robots have been unveiled over the years, in Japan, China, South Korea, the US, and the UK, among others, including by leading automotive companies, such as Toyota. Many have been beautiful designs and engineering benchmarks, but few – if any – have gone beyond event showcases to create a viable market.
Arguably, most of these devices have been little more than expensive shop dummies: window displays to upsell other services and products.
SoftBank Robotics – again known as Alderbaran Robotics, the name of the French company SoftBank acquired in 2012 – has long revealed the problems in the humanoid robot market.
Its toddler-sized NAO humanoid has found a handful of niches in education and entertainment, but the larger Pepper robot struggled to find long-term applications in the public-service roles for which it was designed.
Both robots suffered from a lack of new apps as the hardware became more advanced. Both also revealed a gulf of expectation: a boring and disappointing user experience when compared with 100 years of science-fiction robots. In short, the reality of both devices is primitive functionality, triggered by simple key phrases.
A NAO robot can sing, dance, and tell stories impressively, but ask it a question not covered by its rudimentary onboard programming, and it will say “I do not understand what you want.” The novelty wears off quickly and real utility is hard to find.
New applications can be coded for NAO robots, but again, these will be routines. NAO machines have no onboard AI, and that is the case with most humanoid machines.
Some robots can connect to cloud-based supercomputing or natural language services, such as IBM’s Watson, yet few applications have emerged from that facility.
Amazon has entered the domestic robot realm this year, and recently acquired household device maker iRobot, but Amazon’s own (non-humanoid) robot has hardly set the market alight. With few functions beyond those that could be provided by a smartphone, tablet, or home hub, the advantage of such devices being mobile is unclear, beyond roving home security.
The lesson is simple: the expense is invariably greater than the utility when it comes to domestic machines.
Other humanoid robots have a telepresence or telexistence function, becoming remote avatars for their human controllers, but again, the mass-market utility of that is uncertain.
So, why do robots need arms and legs and to resemble human beings at all? One reason is simple: much of the built human world is designed to be navigated in that way, so some robots will need to adapt to our world more than we will have to exist in theirs.
Arguably, having devices in the future that we can interact with as easily we do with other humans may offer advantages too.
But the engineering challenge is enormous: it is difficult to power a bulky, autonomous, mobile robot that not only contains a powerful onboard computer and sensors, but also dozens of motors, actuators, and servos.
A large onboard battery would be heavy, and so demand more energy to move it around. The usefulness of a robot companion whose battery life is less than an old laptop is far from clear.
Despite all this, there are some promising applications for limbed, anthropomorphic and/or humanoid robots. These include extreme environments, such as space (for which NASA’s Valkyrie and Robonaut machines are being developed), nuclear decommissioning, the energy sector, defence, and security.
Humanoid machines may also help in heavy manual labour in factories and warehouses, as may advanced exoskeletons that can augment human abilities.
But these applications are specialist, well invested, and have years of engineering and coding expertise behind them; they are a world away from the science fiction concept of an intelligent humanoid home help.
And in the latter case, why not just employ a person? Not all friends need to be electric, Mr Musk.