Over the past three years, Peter Fankhauser’s industrial robots have not only been able to climb stairs, but also vault between boxes, perform backflips, and perform other parkour-like tricks.
The robot is not programmed to perform these new actions, but rather adapts to the environment thanks to new artificial intelligence models.
“This is what seems to be the next revolution,” said Fankhauser, CEO of ANYbotics, a Zurich-based robotics startup. “Robots are starting to do really artistic things, and it’s almost scary, because they’re playing with the laws of physics.”
Over the past decade, the $74 billion robotics sector has seen its capabilities accelerate thanks to major leaps in AI, including advances in neural networks and systems that mimic the human brain.
Some of the world’s largest tech and AI companies, including Google, OpenAI and Tesla, are racing to build AI “brains” that can operate robots autonomously, in a move that could transform industries from manufacturing to healthcare.
In particular, improvements in computer vision and spatial reasoning capabilities are enabling robots to gain greater autonomy while navigating a variety of environments, from construction sites to oil rigs to city streets.
Until now, training and programming robots required engineers to hard-code rules and instructions that teach the machines how to behave, often specific to each system and environment.
The recent emergence of deep learning models has enabled experts to train AI software, making machines much more adaptive and responsive to unexpected physical challenges in the real world, and able to learn on their own.
Generative AI, a technology that can generate and analyze multimedia and text, is helping machines better understand the world around them and communicate with humans more easily. This technology now allows people with no coding skills to instruct computers using text or voice prompts.
“It’s like watching a toddler learn,” said Karina Nammi, a partner at London-based early-stage investment fund Plural. “The robots are self-learning, rather than being deterministically programmed, so they don’t have the same high engineering costs.”
Robotics and drones have seen 552 deals totaling $6.5 billion this year, on track to surpass the $9.7 billion raised in 2023.
While many of the advances are expected to be seen in industrial and factory settings, there is also renewed focus among major AI companies on human-like robots, known as humanoids.
Earlier this year, Google DeepMind announced a series of research advances, including using large-scale language models to train humanoid robots to better and more safely understand and navigate their environments. This is also a problem being tackled by World Labs, founded by the “godmother of AI,” Fei-Fei Li, which has reached a $1 billion valuation in just four months.
ChatGPT’s developer, OpenAI, formed a robotics research group last month after dissolving its general-purpose robotics division in 2020 and has also invested in startups, including Figure, which raised $675 million in February from investors including OpenAI, Microsoft, Jeff Bezos and Nvidia at a valuation of $2.6 billion.
OpenAI also invested in Oslo-based 1X Robotics, which raised more than $100 million this year and is working on developing everyday robots for household chores.
The global market for humanoid robots is valued at just over $1 billion, a tiny fraction of the overall robotics market, according to a recent McKinsey analysis, but it’s growing at more than 20 percent annually, three times faster than the market for traditional industrial robots.
Experts warn the technology is still underpowered and expensive — China’s Unitree Robotics sells a humanoid robot for $16,000 — and Elon Musk has said Tesla will start using and manufacturing humanoids next year, and selling them more widely from 2026.
Still, startups are trying to capitalize on the enthusiasm: Investment in robotics and drones is on track to reach $6.5 billion across 552 deals this year, surpassing the $9.7 billion raised across 1,256 deals in all of 2023, according to data from data provider Pitchbook. But total investment in the sector has been steadily declining since 2021.
“Robotics is an incredibly hard problem to solve, and previous solutions have been expensive and inflexible, preventing adoption, especially among small and medium-sized businesses,” said Luciana Lixandre, a partner at venture capital firm Sequoia Capital, which has invested in a number of AI robotics startups, including Robco and Collaborative Robotics. “Advances in AI could help overcome some of the constraints.”
Much of the investment is going to early-stage companies: Warehouse automation robotics company Mytra said this week it had raised $78 million across three funding rounds, while Munich-based RobCo raised $42.5 million in February to support its flexible robot hardware kit.
San Francisco-based Tetsuwan Scientific recently closed a $2.5 million funding round and is looking to create AI robot scientists that can conduct research and physical experiments in the lab. CEO Christian Ponce said the robots can replicate experiments with greater accuracy than humans, allowing scientists to focus on creativity and discovery.
“Generative AI [has been] “You can’t apply it to unimportant, silly things like back-office task processing or accounting software, but applying generative AI to scientific discovery is the most impactful thing we can do,” Ponce says.
Meanwhile, mass consumer adoption of AI tools is having a knock-on effect on attitudes toward robots, said Sonali Fenner of management consultancy Slalom.
This has allowed companies to consider using robots in public-facing environments. Fenner gave the example of a major retail client that deployed Boston Dynamics’ robot dog Spot, powered by Google’s Gemini Pro model, in its stores to evaluate inventory.
“[The hype] “This robot opens the door a little bit as to what is expected in certain environments. Even if you don’t want Spot roaming around your retail store, a robot that’s not too intrusive might be acceptable,” Fenner added.
Ahti Heinla, co-founder of Skype and CEO of delivery-robot startup Starship Technologies, has deployed small grocery robots in more than 100 cities and towns across Europe and the UK, and said he was amazed at how easily people “recognized the robot as a normal participant in public spaces and accepted it as natural”.
