# The race to build robots that can actually think > Companies worldwide are competing to create robot brains that can handle everyday tasks, but the challenge is far harder than anyone expected. - Title: The race to build robots that can actually think · Newsweek - Summary: Companies worldwide are competing to create robot brains that can handle everyday tasks, but the challenge is far harder than anyone expected. Building robot… - Keywords: 1x technologies, skild ai, unitree, physical intelligence, agility robotics, nvidia, humanoid robots, artificial intelligence, robot brains, embodied intelligence, china, united states - Source: Newsweek — https://newsweek.com/global-race-to-build-superintelligent-robots-12441196 - Read time: 3 min - Topics: 1x technologies, skild ai, unitree, physical intelligence, agility robotics, nvidia, humanoid robots, artificial intelligence, robot brains, embodied intelligence, china, united states ## Robots will need to dream 1X Technologies' NEO humanoid robots will process daily experiences like memories, discarding unnecessary data while retaining important details about their owners. > Essentially, robots will dream in the end, because they will need to take all the experiences from today and kind of back propagate it. ## The biggest tech race ever Building robot brains that can think and perform physical tasks is one of humanity's most significant technological competitions, far harder than creating AI language models. ## Hardware existed in 1957 Robot hands could light cigarettes in black and white films from 1957. The hardware was ready decades ago, but the brain has been missing all these years. > The hardware was there even at that time. It's the brain that has been missing all these years. ## China leads in robot bodies Unitree's robots dazzled at China's Lunar New Year gala with choreographed dancing. One Chinese robot ran 100 meters faster than any human in history. ## A man with a controller At Unitree's Hangzhou showroom, the impressive robot movements are actually controlled by a young man in the corner holding a game controller. > The biggest challenge for the whole industry is the brain. All of the industry are trying to develop it. ## Not as smart as squirrels Robots that kick people or strike coworkers in viral videos aren't being malicious. They simply lack intelligence, not even matching a squirrel's smarts yet. > I don't think we've made a robot as smart as a squirrel. ## Walking beats thinking for robots Moravec's paradox explains why hard intellectual tasks like ChatGPT performs need less computing power than easy physical tasks like walking or making coffee. ## Factories become robot schools Industrial sites offer the best environment for robots to learn simple, repetitive tasks. These workplaces are training grounds for developing more advanced intelligence. ## Agility unveils safer humanoids Agility Robotics introduced Digit 5, the first humanoid engineered to work safely alongside humans at scale without traditional physical safety barriers. ## NVIDIA powers the robot race NVIDIA sells three computers for robotics: infrastructure for training brains, testing in simulation, and the robot brain itself. Eighty to ninety percent of global GDP is physical. > We desperately need robots for labor shortages or dangerous jobs, or a combination of robots and humans. ## Compressing 500 million years The human brain evolved over 500 million years since the first neural tissue formed. Robot engineers are trying to compress that timeline dramatically. ## People wear cameras for robots A subindustry has emerged of people with cameras strapped to their heads and wrists performing ordinary tasks to generate training data for robot brains. ## Puppeteering robots to teach them Physical Intelligence collects data by having people marionette robots with two sets of arms, teaching them complicated tasks like folding shirts or cooking meals. > You could watch a video of how to hit a baseball, but then you have to actually go and practice the thing and try it in the real world yourself. ## Renting Airbnbs for robot training Physical Intelligence rented dozens of Airbnbs to help robot brains navigate strange kitchen sinks, different coffee makers, and wiping up spills in unfamiliar homes. ## What robots tell your spouse Robots can share learned skills across networks, but this creates privacy challenges. What your robot discusses with you shouldn't necessarily be shared with your wife. > It's not just that you don't want your neighbor's robot to know. What your robot talks to you about is not necessarily what it should talk to your wife about. ## When the encoder ring falls off Safety isn't just about robots picking up knives instead of oranges. It's about what happens when a part fails, like an encoder ring that tells the robot where its leg is. ## China ahead in hardware only China leads extensively in robot hardware and deployment, but the U.S. remains competitive in developing the brain technology that will ultimately matter most. > The argument becomes China is ahead in many ways because they are doing robotics very extensively, but this narrative is driven by hardware. ## The ChatGPT moment is coming Some experts think the robot brain breakthrough could take decades, but the most optimistic predict a ChatGPT-like moment within just a few years. > If I had written this out before we started Pi, I would have been at maybe five years. Instead, it'll be like two and a half years to get there. ## Key takeaway Building robot brains that can think and perform everyday tasks is much harder than creating AI language models, but companies expect breakthroughs within years.