Robotics won’t have a ChatGPT Moment

The Path to General-Purpose Robots

In 2022, ChatGPT’s launch showed how LLMs can radically transform our approach to working with digital information. Leading talent and $50B+ in capital have since come together to form the AI Revolution, forging the models, tools, and apps of the future. Alongside this digital movement, we are starting to see AI applied to the physical world. Founders and technologists have begun implementing AI into robotics, with tremendous early results, begetting the question of a potential “Robot Moment.” Given the pace of robotics development unlocked by AI, we believe there is a path to achieve general-purpose robots, which will bring robotics into mass-market adoption in our homes, businesses, and industries. Robotics won’t look like a ChatGPT Moment per se, but we do believe advancements in robotics will unlock major productivity gains across the globe and dramatically change daily life related to physical work.

1. Single-purpose robots have existed for 50+ years

Robots have been at work since the 1960s to augment physical labor. Consumer devices like the Roomba have even become popular in our homes. Yet, industrial and consumer robotics have historically grown at a linear pace and not yet achieved exponential adoption. Despite flashy demos, robot projects still encounter challenges operating across different environments and with costs. The robots of today have not figured out how to perform zero-shot, generalizable learning. As form factors like humanoids evolve and robots become general-purpose – adaptive rather than pre-programmed – we believe this technology is on its way to crossing the chasm and entering the mainstream.

2. General-purpose robots are now on the way, but will excel and struggle in ways unrelatable to humans

Robots are surprisingly behind humans in basic capabilities. Generally, single-purpose robots are programmed to perform one specific function well; we see this in industrial use cases across the globe. However, they fail in translating learned behaviors to new tasks and environments or implementing complex reasoning on-the-fly. Unlike other modalities in the digital world, robotics is severely bottlenecked by a lack of quality training data, a major gating factor to achieving general-purpose intelligence. Luckily, open research in the past few years has accelerated progress in generating higher volumes of scalable training data. In tandem, input costs have continued to decline as hardware becomes more affordable. The race is on, with 20+ new humanoids from the likes of Figure and Tesla being developed to break ground on robot generalizability.

3. Robotics won’t have a ChatGPT Moment

Founders, investors, and journalists alike have wondered if a ChatGPT-like event for robotics is coming – a single “WOW!” moment where the latent potential of AI robots is felt in the mainstream. We believe robotics cannot have a ChatGPT Moment due to physical constraints to adoption, high upfront costs of ownership, and the nascency of the ecosystem. Rather, we see robotics crossing the chasm more gradually, as each of us experience our own unique Robot Moments when we interact with robots in a coffee shop or in people’s homes as capabilities mature at a rapid clip.

4. We expect the general-purpose robotics wave to be one of the biggest in technology

We see a thriving robotics ecosystem emerging. Investors have already poured $4B+ into funding next-generation robotics startups over the past decade as leading academic and AV talent form new companies across the stack. As with LLMs, we expect robotics to benefit from tailwinds in accelerating research, accessible compute, and available capital; however, we believe some headwinds like data scarcity, supply chain constraints, and hardware limitations will still pose a formidable challenge. Whether companies take a partnership or vertically-integrated approach, we believe software will drive differentiated value in the ecosystem atop hardware. We are excited about what companies in this layer like Skild and Physical Intelligence (π) can unleash.

If you’re working in robotics, please feel free to reach out with any thoughts and feedback as we look to improve our thinking. We are excited to continue partnering with ambitious founders accelerating the evolution of AI in the physical world. Please reach out to us at robotics@coatue.com if you have a big idea.

Please see important disclaimers.