What Happened
In a recent social media post, Ethan Mollick, a professor at The Wharton School and a prominent voice on AI adoption, framed a critical question about the current technological moment. He posits that whether "this-time-is-actually-different" for robotics is one of the most significant questions for our medium-term future.
Mollick acknowledges the ongoing, profound transformation of white-collar work through generative AI and large language models. However, he draws a crucial distinction: the largest economic shifts in history have occurred in the physical economy. Previous Industrial Revolutions fundamentally changed manufacturing, transportation, and agriculture through mechanization. He argues that for AI to drive a similar scale of economic revolution, it must successfully bridge into the physical world through advanced robotics.
"Yes, white collar work will transform with AI," Mollick writes, "but the physical economy is where previous Industrial Revolutions have occurred & that is where robots will matter."
Context: The AI-to-Robotics Gap
Mollick's comment touches on a central tension in current AI development. The last two years have seen explosive progress in digital, software-based AI—models that write, reason, and create within the confines of a computer. The leap to reliably controlling physical actuators (robotic arms, legs, wheels) in unstructured, real-world environments remains a formidable challenge. Success in this domain would mean AI moving from automating spreadsheet analysis to automating construction, complex manufacturing, logistics in warehouses, and perhaps even eldercare.
The promise is a new wave of productivity not just in information processing, but in the creation, movement, and maintenance of physical goods and infrastructure. The obstacle is that the real world is infinitely more messy and unpredictable than a text prompt.
gentic.news Analysis
Mollick's framing is a necessary corrective to the current narrative, which is overwhelmingly focused on AI's impact on knowledge work. It connects directly to a trend we've been tracking: the accelerating convergence of AI models with robotic platforms. This isn't speculative. In recent months, we've covered significant steps in this direction.
This aligns with our March 2025 coverage of Google's RT-2-X model, which demonstrated improved generalization for robotic manipulation by training on web and robotics data. More recently, in April 2025, we reported on Covariant's RFM-1, a robotics foundation model that uses a diffusion transformer to translate language and video into physical actions, a key architectural step toward more general-purpose robots. These developments suggest the field is moving beyond pre-programmed, single-task machines toward systems that can understand goals and adapt.
Furthermore, Mollick's point about "previous Industrial Revolutions" provides crucial historical context. The First Industrial Revolution was defined by steam and mechanized production; the Second by electricity and mass production. The current digital revolution has so far been largely informational. A true "AI Industrial Revolution" would require the technology to generate tangible, physical output at scale, increasing the productivity of the entire goods economy. The entities pushing this—from Tesla with its Optimus bot to startups like Figure and 1X Technologies—are betting billions that this transition is imminent. Mollick's question cuts to the heart of whether those bets are justified or if robotics will remain a domain of incremental, niche automation for the foreseeable future.
Frequently Asked Questions
What does "this-time-is-actually-different" mean for robotics?
It refers to the long history of hype and disappointment in robotics and artificial intelligence. For decades, predictions of general-purpose robots entering homes and workplaces have failed to materialize, with robots largely confined to controlled environments like factory floors. The phrase questions whether the current wave of AI—particularly large foundation models trained on vast datasets—provides the missing piece (common-sense reasoning, adaptability, and understanding of language) to finally create robots that can operate reliably in the unstructured human world.
How is AI for robotics different from AI for white-collar work?
AI for white-collar work (like ChatGPT or Copilot) operates in the digital realm. Its inputs and outputs are text, code, or images. Mistakes are often low-cost and easily corrected. AI for robotics must perceive and act in the physical, three-dimensional world. It deals with latency, sensor noise, physics, safety, and the potential for high-cost errors (e.g., damaging equipment or causing injury). This makes the problem significantly more complex and requires integrating perception, reasoning, and low-level control.
Who are the main companies working on AI-powered robotics?
The field includes both tech giants and well-funded startups. Google's DeepMind has its Robotics division. Tesla is developing the Optimus humanoid robot. Amazon invests heavily in warehouse robotics. Notable startups include Figure AI (partnered with OpenAI and BMW), 1X Technologies (backed by OpenAI), Boston Dynamics (now focused on commercial applications with Hyundai), and Covariant, which builds AI-powered robotic control systems for logistics. These entities represent the forefront of testing whether Mollick's "different" future is arriving.
