Embodied AI researchers, embodied intelligence is

Embodied AI researchers, embodied intelligence is not a scam. We cannot deny its future development based on past perspectives and current progress. A few years ago, no one thought pure vision-based FSD would succeed, but when it did, its significance to society was immense. FSD is also a form of embodied intelligence, just simpler in difficulty. If FSD can succeed, embodied intelligence can also be realized in the future.

Admittedly, there is a lot of bubble and many "scams" in embodied AI currently. From an investment perspective, choosing the "shovel sellers" is indeed a choice with relatively high certainty. However, I don't think this is a bad thing. Not all investments yield immediate results. Major breakthroughs require significant attention, funding, and talent support. Bubbles can be beneficial. This is also one of the reasons why robotics has developed for decades, but progress has accelerated rapidly in recent years.

In my view, humanoid robots are not limited to emotional value. While affinity is a major advantage, versatility is the most important reason for developing humanoids, and it's the inevitable conclusion from first principles. This is also why I guess Elon wants to build Optimus, not just for internal factory use. The world was created by humans, so the best form to interact with it must be humanoid robots. However, the challenges are indeed huge. Whether it's hardware, algorithms, or data, everything is currently limited, and it requires massive capital investment. It's actually quite similar to the emergence of SpaceX's reusable rockets. So-called "impossible" things or things the general public "doesn't recognize," as long as they are not limited by physical laws and are meaningful, are worth encouraging. In my view, humanoid robots and embodied intelligence are not constrained by physical laws; they just need time. The US is lucky to have Musk, and the world is lucky to have Musk.

Imagine a future where the streets are filled with all kinds of robots, humans have begun interstellar travel, with nearly infinite energy and productivity. Such a future is still very much worth looking forward to.

$Tesla(TSLA.US)$SpaceX(SPCX.US)

吉姆哈克的交易员

Here's a bold statement: Embodied AI worldwide is a scam.

Or to put it another way, robots are a good thing, but embodied AI is a pure scam. When it comes to long-term tasks, currently zero intelligent agents can complete a truly "autonomous" long-term task.

Being able to move, being able to complete tasks, and being able to make money are three different things. The typical embodied AI product launch only proves the first one, but by the time the valuation is set, the third one has already been completely overhyped.

Humanoid form is not the optimal solution for factories; it's the optimal solution for emotional value. Factories want cheap, stable, and maintainable, not something that walks on two legs and can dance on the Spring Festival Gala.

The most amusing thing about companies going public now is that they package robots as the physical embodiment of AGI to raise funds. If software makes one mistake, users complain once; if hardware slips once, it's an accident. The fault tolerance is worlds apart.

This kind of industry itself should be avoided, and even more so, companies that treat government demonstration orders as recurring revenue. So I often tell my friends, if you want to chase a concept, you might as well buy companies in precision transmission, force control sensors, and the like. Selling shovels is closer to cash flow than selling dreams.