We are firmly in Phase 3 of AI, at least from a communications point of view. The term Physical AI is now seen all over social media, YouTube, financial media, and manufacturing media. There are certain groups that are very excited about this.
What is it, exactly?
Let’s start with the first two phases. First was Generative AI; you and I are most accustomed to that. It’s the ChatGPT, Gemini, Claude, and Copilot type of interaction, offered respectively by OpenAI, Google, Anthropic, and Microsoft. It’s where you can toss ideas around with a digital partner, you can research, you can have a digital conversation about any topic, you can create business plans or just content in the form of text, graphics, and videos. You can also have it write code if you have the right Generative AI package.
Then came Agentic AI, which goes a step further than Generative AI, because like Generative, Agentic AI’s diet is mostly data. However, Agentic AI was built to make sense of data, to let data fuel decisions and then act on those decisions. As if a company “agent,” this type of AI makes inroads into acting like an employee, in the sense that it takes input, figures out a plan (or pulls from different action alternatives), then acts. Still, it lives in the world of data.
That’s where Physical AI is so different. As its name implies, it has to do with the physical world. So, this type of AI is responsible for many of the functions of self-driving cars, Autonomous Mobile Robots (AMRs), and driverless forklifts. It can use sensory input from visual, auditory, touch, and smell, in a way. It gathers information not from a database but from the world around it. Even a dynamic world is fair game for this type of intelligence, where a sensor-equipped AMR can dodge a human being, another AMR, a forklift, etc.
Physical AI is not new. One of the early stories on this site and newsletter involved Mitsubishi and its development of this type of AI, based on its experience in the autonomous automobile market (see the story from March 2021 here: https://fifthwavemfg.com/mistubishi-electrics-maisart-artificial-intelligence-meets-mc-machinery-laser-cutting-and-edm/).
Much of the attention today is focused on the robotics market. You should have some focus on this too, because like manufacturing, this is focused on doing, not being or just thinking. Only yesterday it was announced that robotics provider RobCo, based in Munich and with offices in San Francisco and Austin, TX, is receiving a $100 million USD Series C investment from a group led by Lightspeed and Lingotto to roll out its technology and greatly expand its footprint in the U.S. market.
It’s been a banner week for investments—two days prior, Vention (Montreal) gained an investment of $110 million USD from a group of companies that includes major Canadian investors as well as the NVIDIA venture capital arm. Vention will use the investment to accelerate its own Physical AI research, bring new capabilities to its software platform, expand pre-engineered applications, and strengthen its global footprint, concentrating on Europe, Middle East, and Asia. Its own vision of the near future is a set of automation offerings that fit together without separate, manual integration. This is a great goal and good luck to both of these companies. (See our brief interview of Vention Founder and CEO Etienne Lacroix as well as a whirlwind Automate booth tour here: https://www.youtube.com/watch?v=d3Siz41Yi6k.)
I can hardly wait to see what comes next.