Futuristic, humanoid companions were fun to think about in movies starring Harrison Ford but less so in reality. A Guardian article from 2003 went as far as describing robots of the day as a “motley crew of motorized bins.”
It’s a sentiment that might soon feel out of date if robot enthusiasts have their way.
Following AI’s breakout moment last year – driven by large language models (LLMs) powering chatbots – roboticists have been in desperate search of a breakthrough that will give them their ChatGPT moment.
“I’ve been asked what’s the biggest thing in 2024 other than LLMs. It’s Robotics. Period,” Nvidia’s senior AI scientist, Jim Fan, wrote on X in December. “We are ~3 years away from the ChatGPT moment for physical AI agents.”
Fan noted that the robotics industry has been “cursed” for too long by the Moravec’s paradox – the observation that stuff humans can easily do is much tougher for robots.
Some recent breakthroughs suggest that curse may be lifting, putting the robotics industry on a course that could edge it towards a ChatGPT-like breakthrough.
The robots are coming
Elon Musk got enthusiasts hyped up with the introduction of a next-gen version of Tesla’s Optimus robot in December.
The robot appears to be capable of doing exercise, as well as other delicate tasks such as picking up an egg.
Last week, Google DeepMind introduced a bunch of innovations that might make the prospect of living alongside robots in your home safer and more appealing.
The most important one was a feature called Auto RT. This harnesses the power of LLMs and other AI models, like vision-language models (VLMs), to effectively give robots chore-busting powers, for example.
As DeepMind puts it, a robot working with this system could feasibly collect data on its environment through observation, before generating a to-do list of tasks that it could carry out in response to, say, a messy kitchen.
“For each robot, the system uses a VLM to understand its environment and the objects within sight,” the company wrote in a blog post. “Next, an LLM suggests a list of creative tasks that the robot could carry out.”
On Sunday, AI robotics company Figure, based in Sunnyvale, announced that its robot Figure-01 had figured out how to make coffee. “Our AI learned this after watching humans make coffee,” founder Brett Adcock wrote on X.
A video posted by Adcock shows a man asking Figure-01 – a metallic, i-robot-like humanoid – to make a cup of coffee with a Keurig coffee machine. Within a matter of seconds, it flips open the machine’s lid before placing a coffee pod in it to get a brew going. Check it out here:
Robots were in the spotlight at this year’s CES in Las Vegas too. Though prototypes are often shown off at the event, this year’s robotics display has sought to highlight how AI can make robots useful.
Nvidia’s vice-president of robotics, Deepu Talla, spoke at length during an address ahead of the event about how some of its partners – like Boston Dynamics and Collaborative Robotics – are using generative AI to help robots better “understand text or voice commands.”
Samsung also showed off a new version of Ballie, a robot it first unveiled at CES in 2020. A video shown of the yellow, Roomba-sized AI robot shows its smart device controls in action at home, as it switches on lights, serves as a companion during workouts, and even makes phone calls for you.
AI alone won’t make the path forward for robots straightforward, however.
Getting robots to function well has been a major challenge for some of the most well-resourced companies over the years.
As my colleague Eugene Kim previously reported, Amazon’s home robot Astro has struggled to get traction. It has faced scathing reviews and been dismissed internally by sceptics.
That said, Amazon is now turning to a new AI model to boost the robot and make it a smarter companion.
Success could make it and others big moneymakers in the future – even if they aren’t right now.
Research from financial firm Macquarie, published in 2022, emphasized the importance of home-based demand for humanoid robots.
If they’re any good and start being bought by consumers, it could tilt the industry towards a $3 trillion market by 2050, the firm’s analysts think.
If the future can deliver robots that are actually helpful, the robots might not be so lame in the end.
Read the full article here
