Build an autonomous self-replicating robot. Not a RepRap—those take hours of careful human labor to build. Design an assembly robot to put together parts: something that can pick up a screw, place it in a hole, and screw it in place. This is a hard problem (some current PhD students do similar projects for their research) but the required tools all exist, including ROS and existing robot arm control routines. Some challenging coding is required to execute the assembly steps in order, and some challenging systems engineering is needed to make sure the robot has the observation and control accuracy capabilities needed (e.g. hobby servos may not be accurate enough). The robot arm must itself uses only off-the-shelf parts (stepper motors, encoders, laser-cut parts, off-the-shelf shafts, etc.). This would be a hard problem to solve, but is solvable by someone who’s great at coding, mechanical engineering, and systems engineering.
The munchkin part occurs when this is built. Existing robot arms typically cost ~2x or more the cost of the parts (so the manufacturer can pay assembly labor cost, engineering overhead, etc.). This autonomous assembly robot could thus be substantially cheaper than conventionally manufactured assembly robots. And, scaling up production does not require more workers, just more table/shop space for more robots. So, one robot could be used to make a very large number of additional robots, resulting in a large profit for the owner.
The munchkinnery continues by enabling full automation of more challenging manufacturing problems. Assembling a robot from parts purchased online is easier than, say, assembling a factory. To compete, people would race to automate increasingly complicated problems. Once one person had automated a task (requiring a large investment), others could copy the control code and designs for free (or a license fee) and enjoy increased productivity with lower costs. This would make manufactured goods much cheaper, and possibly make recycling and Mars colonization cost-effective.
Data to support the feasibility of this includes existing examples of high-level robot tasks (welding robots for cargo ship fabrication, navigation and mapping of buildings by drones, various capabilities of the PR2 robot platform) and NASA studies of the feasibility of moon colonization using self-replicating robots.
Side effects may include massive unemployment, creation of a robot-administered police state, and increased risk of self-improving AI.
Build an autonomous self-replicating robot. Not a RepRap—those take hours of careful human labor to build. Design an assembly robot to put together parts: something that can pick up a screw, place it in a hole, and screw it in place. This is a hard problem (some current PhD students do similar projects for their research) but the required tools all exist, including ROS and existing robot arm control routines. Some challenging coding is required to execute the assembly steps in order, and some challenging systems engineering is needed to make sure the robot has the observation and control accuracy capabilities needed (e.g. hobby servos may not be accurate enough). The robot arm must itself uses only off-the-shelf parts (stepper motors, encoders, laser-cut parts, off-the-shelf shafts, etc.). This would be a hard problem to solve, but is solvable by someone who’s great at coding, mechanical engineering, and systems engineering.
The munchkin part occurs when this is built. Existing robot arms typically cost ~2x or more the cost of the parts (so the manufacturer can pay assembly labor cost, engineering overhead, etc.). This autonomous assembly robot could thus be substantially cheaper than conventionally manufactured assembly robots. And, scaling up production does not require more workers, just more table/shop space for more robots. So, one robot could be used to make a very large number of additional robots, resulting in a large profit for the owner.
The munchkinnery continues by enabling full automation of more challenging manufacturing problems. Assembling a robot from parts purchased online is easier than, say, assembling a factory. To compete, people would race to automate increasingly complicated problems. Once one person had automated a task (requiring a large investment), others could copy the control code and designs for free (or a license fee) and enjoy increased productivity with lower costs. This would make manufactured goods much cheaper, and possibly make recycling and Mars colonization cost-effective.
Data to support the feasibility of this includes existing examples of high-level robot tasks (welding robots for cargo ship fabrication, navigation and mapping of buildings by drones, various capabilities of the PR2 robot platform) and NASA studies of the feasibility of moon colonization using self-replicating robots.
Side effects may include massive unemployment, creation of a robot-administered police state, and increased risk of self-improving AI.