[Repost] Physical AI is a scam.
Translated by AIView original
Source: https://ianpark.vc/p/physicalai/
A. Technical limits (Can it work?)
- LLM success does not carry over to the physical world as is
- There is no guarantee that throwing more money and time at it will work
- Robots don't have the internet's data
- Current models are weak on the long tail: Tesla and Waymo
- Robots don't improve at the speed of software
B. Limits of commercialization (Can it become a real business?)
- AI could also make humans cheaper
- An AI's mistake is a human's life
- Physical AI is a multiplication problem, not an addition problem
C. Technology success vs. investment success (Can we make money?)
Why humanoids are Clean Tech 2.0
They look astonishingly similar to today's humanoids.
- First, it was an industry of atoms, not bits. Clean tech couldn't be copied and distributed like software. It needed factories, materials, equipment, supply chains, and certification.
- Second, there was a commercialization valley of death. Clean tech succeeded in demos, but cost, yield, and reliability broke down at mass production.
- Third, it had to compete with existing systems that were already cheap and mature. Clean tech's competitor was fossil fuels, and now it's human labor. Even if human labor has limits, that doesn't automatically make robots more economical.
- Fourth, narrative and FOMO got ahead of economics. Clean tech was also a "technology that will change the world," but that fact doesn't guarantee the conclusion that "you have to invest at this price."
- Fifth, the winners of the technology and the winners of the investment were different. Solar became a huge industry, but early investors didn't take the biggest returns. Companies with longer-lasting capital and manufacturing capability captured most of the value.
The lesson from Clean Tech 1.0 is clear.
The technology was right but the timing was wrong, and the industry grew but early investors were not necessarily the winners.
So what? (So what?)
Don't get swept up in FOMO, look at the essence
So when do you invest in robots?
The day I can close my eyes and sleep in the driver's seat while my Tesla drives fully autonomously from my home in San Francisco to my relatives' house in LA. I will use that day as the starting point for investing in Physical AI.
1 comment14 views
Comments 1
Translated by AI·View originals
Relatable.