The Future Can’t Be Built by Mimicking the Past. Right. Right?
Lily, our robot, once tried to go downstairs and ran into… a door 🧐.
Some say the fix is obvious: give Lily a pair of humanoid arms just to twist the knob. Maybe. But maybe the simpler answer is retrofit the door to be ADA compliant, add an API so it opens automatically, or, brace yourself, just ask a person to help with the seldom-used door.
That little incident highlights a bigger opportunity: it is good practice to stop and question all assumptions. Two assumptions that stand out to me are:
That our past is somehow better than our present or future.
That humanoid robots are the inevitable form factor.
The tech industry seems to be making the same assumption about humanoids: if the world is built for humans, then robots should simply take the human form factor. We put Lily into a real environment and quickly realized that this assumption might be faulty. Some of the humanoids being demoed today look like progress, but often resemble complicated solutions searching for old problems. Mimicking the human body might be optimizing for a past state, not for the world we are moving into.
This does not mean humanoid work is wasted. Far from it. Advances in motion control, reinforcement learning for gait, and dynamic balance are incredible, and they will transfer into many specialized systems. That is progress worth celebrating. But none of it makes humanoids inevitable.
Assumptions are necessary. But holding them as absolute, un-impeachable truths is counter productive. Sure, your v0 or v1 product might be built on clearly articulated assumptions — maybe even a TAM of $1 Bajillions.
But once you have real feedback from your last releases, you have to decide what game you are playing.
If your goal is to build something truly useful and durable, then you must be willing to abandon those a priori assumptions. If your goal is to raise a boatload of cash so that you can raise another round, then forget I said anything 😃.
As Robert Sun put it at Actuate:
“Reframing debate around form factor, instead of merely recreating ourselves in humanoid form, we must ask: if we are the architects of new workforce, what is optimal form for the complex jobs we need done?”
So here’s the question: is the human form factor — speech, vision, locomotion — or the past itself really a good enough blueprint?