Why Are Robots Waiting for Their Own "ChatGPT" Moment?

Since ChatGPT emerged in late 2022, the technology landscape has changed dramatically, moving artificial intelligence from the realm of promises and hype into a daily reality used by millions. Yet the field of robotics, despite decades of research and development, has not yet witnessed its own comparable turning point. So what stands in the way of this breakthrough? This is the question posed by Les Karpas, the global head of physical AI at NVIDIA's Inception program.

A Turning Point Not Yet Reached

Karpas believes that robots are waiting for their own moment, similar to what ChatGPT achieved for large language models. While NVIDIA has been highly enthusiastic about this sector, as evident in the remarks of its CEO Jensen Huang over recent years, there remains a fundamental obstacle preventing the industry from reaching that decisive point that would move robots into widespread use.

The Data Problem at the Heart of the Challenge

Karpas sums up the core obstacle in a way that is easy to understand but extremely complex to solve: no one possesses a massive, internet-scale database dedicated to physical AI, of the kind that has enabled companies such as OpenAI and Anthropic to train their language models on the enormous volume of text available on the web.

Even in the field of autonomous driving, companies like Waymo and its competitors have managed to build their databases through years of accumulated miles driven on roads, databases that continue to expand market after market. General-purpose robots, however, lack such a natural reservoir of data.

Emerging Solutions to Bridge the Gap

Faced with this shortage, a growing ecosystem of startups is seeking to generate this vast volume of data through synthetic means, relying on several innovative approaches, most notably:

• Digital simulation of real-world environments.
• Generating synthetic data to train models.
• Building foundation models trained on multiple forms of robots simultaneously.

The challenge of bridging the gap between the digital and physical worlds represents one of the most complex issues facing the robotics industry today, which makes providing the right data the cornerstone of any future progress.

Multidisciplinary Expertise

Karpas is among the most qualified people to answer this question, given his position, which allows him to coordinate with a broad ecosystem of startups spanning robotics, automotive, manufacturing, mobility, smart cities, and more—the very fields attempting to address the data gap.

His professional career reflects the nature of this industry, which demands multidisciplinary skills. He has moved between various roles, including working as an architect, a manufacturing engineer, a startup CEO, and an institutional venture capital investor, across well-known organizations in different sectors.

Conclusion

Karpas's insight reveals that the future of robotics does not hinge on developing more sophisticated hardware so much as on solving the data dilemma that enables these machines to understand the physical world and interact with it flexibly. Once this dilemma is solved, we may witness the long-awaited robotics turning point, just as ChatGPT transformed our relationship with text and language.

✦ بقلم فريق دروب أيديا