The Case for Common Sense in the Age of Physical AI

Physical AI has gained significant attention, especially through presentations like Jensen Huang's at GTC 2025. While often linked to robotics, its applications span:

• Power

• Water

• Waste

• Transportation

• Manufacturing

• Healthcare

• Agriculture

• Information technology infrastructure. 

In these sectors, Physical AI helps achieve results beyond what domain knowledge alone can offer, though it's unnecessary when domain knowledge suffices.

Why Did the Wine Glass Shatter in the Memorex Commercial?

Despite the attention surrounding AI, many issues can be addressed with classical engineering knowledge and basic tools. In my experience, we can solve most problems efficiently using formulas, experience, and low energy-consuming tools such as calculators and laptops with basic and advanced analysis software.

Consider the following example.

In the 1970s, Memorex Corporation aired a television commercial demonstrating that Ella Fitzgerald could shatter a wine glass with her singing. The commercial showed that when her voice was recorded on a Memorex tape and played back, it shattered a wine glass as well. 

The advertisement concluded with the line, "Is it Real, or is it Memorex?"

So, why did the wine glass shatter? 

Basic domain knowledge suggests that the prominent frequency component in the singing coincided with the natural frequency of vibration of the wine glass. This channeled the energy into a state of resonance. At resonance, the high amplitude of vibration of the glass caused the breakage.

Based on domain knowledge, one would also conclude that the wine glasses for both cases— live and playback on tape—had to be of the same geometry and material. Furthermore, the glasses also had to be placed under the same boundary conditions. A key boundary condition is the state of placement on the table, e.g. placed freely on the table the same way in both cases—live and on tape. 

If the wine glass was placed freely on the table in one case and rigidly in the other, its natural frequency may not coincide with the input frequency in the singing, and it might not break. Understanding this boundary condition is important.

What can we learn from the shattered wine glass?

The wine glass demonstration is an excellent example of many real-world vibration-related issues an engineer might encounter. Rotating systems, such as fans, washing machines, turbines, hard disc drives, and vehicles, are a few examples. Resonance induced vibrations can be detrimental in these systems. 

Undergraduate courses in physics and mechanical engineering have taught us to look for offending frequencies. For example, when a washing machine starts to vibrate excessively, one may redistribute the load to see if that fixes the problem.

To my dismay, in my conversations about the example, the first reaction to explaining the wine glass root cause phenomenon is often “collect data and infer.” More recently, with the rise of Generative AI, the reaction is to type this question into a generative AI tool

Even Good GenAI Outcomes Require Domain-specific Prompts.

I decided to try out GenAI to illustrate that even GenAI requires domain-specific prompts to get useful outcomes. The following was the response to my initial question posed to Microsoft Copilot.

Why did the wine glass shatter when Ella Fitzgerald sang in the Memorex commercial on TV?

What are the shortcomings of the prompt, and how can we refine it?

The Copilot answer, while citing resonance, does not explain that to replicate the effect, the physical design, material, and boundary conditions for the wine glasses would have to be the same. For example, if the wine glass is free during live singing and clamped to the table when the tape is played back, it may not shatter as its natural frequency would have shifted up to a higher value.

Therefore, the prompt needs more domain-specific steering. So, I next created the following input.

I then followed this up with a more refined domain-specific prompt.

The simple answer that I was expecting was that the natural frequency of the wine glass would shift. While I got a reasonable answer, it was still not sufficient. It also shows that my prompt needed to be better.

Why Use a Chainsaw to Cut Butter?

During my 42 years of experience in systems engineering and the application of physical AI, I have learned to always seek the least-energy solution. So, for a physical system, simple or complex, I first apply my domain knowledge and use basic tools to solve a problem.

In the wine glass example—which is a good representation of many a vibration problem I have encountered in complex physical systems—the use of AI tools is tantamount to using a chainsaw to cut butter. 

A least-energy outcome is simply using domain knowledge.

About the Author

Former SVP, Chief Engineer, and Senior Fellow at HP, Chandrakant is a leader in AI, energy-efficient computing, and sustainability. He is an IEEE Fellow, ASME Fellow, member of the NAE, and the Silicon Valley Engineering Hall of Fame. Follow me on LinkedIn or email me at [email protected].

Want to stay ahead of the curve with insights into the newest advancements in Physical AI? Subscribe to Chandrakant’s newsletter at GenAI Works.

🚀 Boost your business with us—advertise where 10M+ AI leaders engage

🌟 Sign up for the first AI Hub in the world.

📲 Our Socials