Here are some ideas for robotics startups.

Present-Day Ideas

This section is for ideas which I think are feasible to build using current robotics techniques (building maps, using explicit poses, grasping - the “NVIDIA” robotics approach, if you will).

Grocery

Supply Chains

Waste Management

Medical

Home

Home-Adjacent Services

Old Age

Some other problems that the elederly face which might be solvable using robotics:

Future Ideas

Given the current state of machine learning, the big upcoming wave of robotics ideas will likely be systems which can understand very open-ended environments and tasks, and deal with large amounts of ambiguity. Consider a system like ChatGPT, but trained using representations of the physical environment in addition to large-scale internet text.

So, to summarize the essential requirements for a startup to compete in five to ten years:

  1. Low hardware cost
  2. Large form factor
  3. General purpose
    • General mobility: Multiple legs
    • General manipulation: Multiple arms

Quadroped

A Spot-like quadroped robot, with an arm attached, could be relatively cheap to manufacture and would be able to do a lot of useful tasks, such as:

Additionally, the extra stability provided by four legs would allow it to carry a larger battery and onboard computer, enabling it to do more sophisticated world understanding.

Biped

Tesla’s robot looks to be a pretty cool platform for interacting with the world. The question is, could a startup manufacture a bipedal robot which would be a better value proposition than Tesla’s? I think most people would probably say no, but here are some arguments in favor of the startup: