èƵ

Don’t be fooled by Elon Musk’s chatty Optimus robots

The long history of robotics should teach us to be more sceptical when it comes to autonomous humanoid robots, says Nicole Kobie

Optimus robots wandered around the party held after Elon Musk’s Tesla unveiled its robotaxi last month, doling out drinks and chatting with guests. They were also being photographed, as android butlers make for great social media content. Partygoers couldn’t believe their eyes – and they shouldn’t have. The robots weren’t fully autonomous, but remote-controlled avatars.

That shouldn’t have surprised the tech-savvy revellers. After all, when Musk first unveiled Optimus in 2021, the humanoid robot that strode onto stage was actually a costumed dancer. Indeed, throughout the long history of robots, if one impresses with its intelligence, it’s a safe bet it is an avatar or carefully programmed.
The very first humanoid robots were hulking metal androids – and used similar tricks to those at Musk’s party. Eric was made in 1928 by William Richards to deliver a speech at the Exhibition of the Society of Model Engineers. It couldn’t walk, but could gesticulate and turn its head.

Most impressively, Eric could converse, and was interviewed by journalists. Richards didn’t reveal how, but implied the magic was managed via radio: Eric was less a metal ancestor to Alexa and more an overengineered walkie-talkie.

Subsequent robots have gained further skills, notably in the late 1960s, when the research institute now known as SRI International built the remarkable Shakey, which could navigate and make decisions. More recently, Boston Dynamics helped develop robots that walk, and though it prefers four-legged designs, a military contract sparked the creation of the bipedal Atlas, its humanoid robot that climbs, flips and runs in videos, but is only just starting to shift into commercial applications.

Boston Dynamics’ robots have embedded intelligence that lets them walk and dodge obstacles without human help, but even they aren’t fully autonomous. In 2020, the company posted a video of Atlas dancing to the song Do You Love Me? – but that took three months of programming.

Autonomous humanoid robots have been built, but remain limited. The best known is perhaps Pepper, developed by Aldebaran after it was acquired by SoftBank. It glides on wheels, its smooth, white, plastic form mimicking the curves of a nipped-in waist and skirt, a control tablet on the front. Released in 2014, Pepper made headlines, finding work as a receptionist, even reading funeral rites. But it failed to find purpose – it couldn’t empty your dishwasher or use a vacuum cleaner – so production was halted in 2021.

Start-ups like Figure and Agility continue the work towards a general-purpose humanoid robot, but arguably we don’t need robots that look like us. Legs may be more useful than wheels to move around without replacing every staircase with ramps, but four is easier for balance than two, hence “dog” style robots like Boston Dynamics’ Spot, already in use for monitoring and maintenance.

Indeed, there were more than 4 million s in use globally as of 2023 and millions more robots autonomously vacuum our floors. Yet they aren’t shaped like us: why automate a human to hoover your home when a Roomba is cheaper and needs less space in the cupboard?

One day, we may have the choice: home automation through a network of smart appliances or a single robot butler. But in the meantime, when a humanoid robot struts up for a chat and a drink, by all means be impressed – but, until proven otherwise, assume it is an avatar or is heavily programmed. When a general-purpose, autonomous, bipedal robot truly does arrive, tech firms won’t be shy about letting us know. After all, Musk held a press conference to announce Tesla was starting work on robotics. Imagine the theatrics if his firm actually produces a real robot butler.

Nicole Kobie is a journalist and the author of

Topics: robotics / Robots / Technology