Is the substantial investment in teaching robots 'mundane tasks' like clothes folding truly justified by its return on investment?

Is it reasonable to invest heavily in teaching robots trivial tasks like folding clothes, in terms of return on investment?

From a short-term and single-task perspective, it's extremely unreasonable. However, from a long-term and technological development perspective, this investment could be incredibly profitable.

This indeed sounds a bit like "using a sledgehammer to crack a nut"—spending millions or even tens of millions of dollars just to solve a household chore that we can handle in a few minutes. If the sole goal is "to fold this piece of clothing," then the current return on investment is definitely negative.

But the key issue is that, for robots, the task of folding clothes is an "ultimate test."

You can think of it like getting a driver's license. The purpose of getting a driver's license is to drive on the road, not to perfectly "parallel park" in the driving school. "Folding clothes" is that "parallel parking"—it doesn't create immense value in itself, but it proves you've mastered a set of extremely complex general skills.

For robots to learn to fold clothes, scientists need to tackle some of the toughest challenges in robotics:

Visual Recognition (The eyes need to understand): The robot must distinguish between a T-shirt, pants, and socks from a pile of messy clothes. Clothes are soft and deformable, which is infinitely more difficult than recognizing a fixed cup or a book. Solving this problem will enable robots to identify and locate objects in various cluttered environments.
Dexterous Manipulation (The hands need to be skillful enough): Folding clothes requires a series of delicate actions like pinching, pulling, lifting, and folding. Human hands can easily perceive the texture and force on fabric, but this is very difficult for robots. Developing "dexterous hands" capable of handling soft, deformable objects is a holy grail in robotics. Once successful, these hands could not only fold clothes but also perform surgery, assemble precision instruments, care for the elderly, and much more.
Decision Planning (The brain needs to figure it out): The robot needs to plan its own steps: "first flatten, then fold the sleeves, then fold in half..." This process requires it to understand the task and continuously adjust its actions based on the real-time state of the clothing. Behind this are complex AI algorithms.

Therefore, the real logic is this:

Short-term goal: Teach robots to fold clothes.
True purpose: By conquering the comprehensive challenge of "folding clothes," illuminate an entire tech tree of general-purpose technologies (advanced vision + dexterous manipulation + intelligent planning).

Once this set of technologies matures, its application scenarios will be vast. A robot that can perfectly fold clothes, with minor modifications, could become:

A caring attendant in nursing homes: capable of feeding, dressing, and tidying beds for the elderly.
A super blue-collar worker in factories: able to complete various complex, non-standard assembly tasks.
An all-around assistant in laboratories: capable of operating various precision experimental equipment.

This is akin to when we invested heavily in computer research; initially, they could only perform simple mathematical calculations, were bulky and expensive, far less efficient than an accountant using an abacus. But who could have imagined that this technology would eventually usher in the entire Information Age?

Conclusion:

What we see as "folding clothes" is actually a crucial stepping stone towards the ultimate goal of "general-purpose humanoid robots." Investors aren't betting on the "folding clothes" function itself, but rather on the immense, multi-trillion-dollar market that will open up once this technology achieves a breakthrough.

In the short term, the ROI seems ridiculously low; but in the long term, this could be paving the way for the next industrial revolution and a transformation in our way of life.