ANA Avatar XPRIZE

I don't take any issue with the sentiment; they want to advance specifically commercial domains by incentivizing them to accomplish moonshots.

The problem I have is that the moonshots (quite literally in the case of the Google Lunar XPRIZE) are so ambitious that the only people that have the resources to even come close to accomplishing these goals would never be interested in prize amounts this small. They would almost definitely end up losing money with the purses the size that they are and quite honestly winning an XPRIZE just doesn't carry enough prestige to counter the capital losses.

The amount of R&D needed to get a robotic avatar to the level proposed here is staggeringly larger than $10mil.

I just don't see the point.

Take pg 7 of the draft guidelines: https://avatar.xprize.org/sites/default/files/ana_avatar_dra...

In the 'Vision' section it requires that the Avatar be able to track motion: "Can you correctly follow an object moving on the floor in front of you?"

The hard way to do this is to develop a vision system with neural nets to track relative motion of any object within the field of view. Objects need to be classified, identified and tracked. Further, the purpose of the object must be divined. That is extremely difficult.

But, if you allow for a wireless connection with a human in a suit and the human operator's brain is doing all of that work (the work a child does when catching a ball), then the challenge gets much much easier. Pass the data to the human (visual) in close enough to realtime so that the human can react while wearing a control suit. Then pass the control suit reactions back to the robot. The problem becomes one of data compression, latency in the system, UI/UX design on the human side of things, etc.

The point is, cheaper, less comprehensive systems could be built that could functionally complete the challenge and win the prize. And those systems could have immediate uses.

For example, what if I have elderly parents who have a robotic assistant at their home fulltime. Some tasks are easy to program (ex. changing a bedpan). But other tasks are much more difficult: a sponge bath to prevent bed sores. The robot could change the bedpan whenever it needed to automatically like a roomba doing the vacuuming. But when it needed to do the bath, it could call me up at home, I could put on my control suit, connect over the internet, and bath my parent from thousands of miles away. The complex 'thinking' and task management of the robot is handled in my brain.

But the basic business case is met: work can be done by a robot with a human 'guide' or 'pilot' at a distance.

Great coincidence