The worst challenges to having the robots we want are non-technological.

Let’s face it, everyone loves robots. They’ve been indispensable as a science-fiction mainstay longer than most folks have been alive. Now, tantalizingly, is the possibility that real robots, the kind we all want, will become a reality, but there are challenges, the worst being non-technological.

In the 60s and 70s we all thought we’d be traveling to work in flying cars and living on the Moon by 2001. That didn’t happen, not because those things were not technologically achievable, but because we did something else. Two generations of humans took a 40-year hiatus to build the internet, including the underlying technologies that make it all work.

Recently, hardware in the form of rockets, robots, and autonomous cars has seen a resurgence in interest as investors began to wonder if the best fruit was disappearing from the tree of software. Briefly, robots had a chance, but now there is AI. Many people imagine AI would enable robotics, but this is not the dynamic we are seeing. AI is a new software tree promising a new generation of low hanging fruit that can endow low venture investments with the possibility of dramatic returns.

For roboticists, it is on us to figure out how to make robots less costly. It is not sufficient to do exciting things—we must learn to do them cheaply enough that we can attract the capital we need to bring them to market.

A tough time is ahead for robotics. Investors will be distracted by the opportunities offered by pure-software AI companies. It is on us to re-envision robotics in a way that can evolve and grow, so we can have the robots we dream of to help make our lives better, free the world from boring labor, and—if we’re lucky—lead humanity into an era of thought, imagination, and creativity.