AI News

OpenAI disbands independent robotics team, researchers merged into general model team

ZhidongxiUpdated Nov 15 04:27

AI summary

According to Fortune, OpenAI has disbanded its independent robotics team and merged the relevant personnel into other teams; the company confirmed this adjustment to Fortune. The robotics research team will be merged into the main research team and report to Chief Research Officer Mark Chen, while the applied robotics team and hardware team will be merged into the Applied division, overseen by President Greg Brockman. OpenAI said the move is intended to better advance robotics research rather than to exit the field, and mentioned that general frontier models, including GPT-6 Astra, are beginning to demonstrate capabilities that can be used directly for robotics.

Why read this

AI score 64

The original article outlines the timeline of two adjustments to OpenAI's robotics team and provides the roadmap and cost background after the organization was merged into the general model system.

Topics

Tags

Body · AI translation

Robot Frontier reported on November 14 that, according to Fortune magazine, OpenAI has disbanded its independent robotics team and merged the relevant researchers into other teams.

OpenAI confirmed this adjustment to Fortune. A company spokesperson said that robotics remains a core component of its long-term vision, and that this reorganization is intended to better advance robotics research rather than to exit the field.

The company disclosed that its robotics research team will be merged into the company's main research team, reporting to Chief Research Officer Mark Chen; the applied robotics team and the hardware team will be merged into the larger Applied division, overseen by OpenAI President Greg Brockman.

There had been almost no public indication of this adjustment beforehand. OpenAI had booked a booth for this week's top robotics conference CoRL in Austin, and its location was marked on the conference app and the booth floor plan.

But when a Fortune reporter arrived at Booth 208, there was no one there, and no OpenAI logo was in sight. On X, some netizens likewise expressed confusion over OpenAI's absence.

I. Within Five Years, the Robotics Team Has Been Adjusted Twice

This is not the first time OpenAI has adjusted its robotics business.

OpenAI had a robotics team in its early days, but gradually shut it down in 2020. Co-founder Wojciech Zaremba later explained that robotics data was limited at the time, and that AGI could be achieved more efficiently through other technological paths.

In 2024, OpenAI re-established an independent robotics team and announced a partnership with the startup Figure to develop AI models for humanoid robots. In February 2025, the two sides terminated the partnership, and Figure said it would develop its own robot models.

Currently, OpenAI is still developing its own humanoid robot. CEO Sam Altman's goal is to hold the first public demonstration in 2027. Yet five months ago, he had publicly recruited for the robotics team.

The spokesperson said that despite the reorganization, OpenAI remains committed to bringing general-purpose robots into the real world.

OpenAI engineer David Schnurr responded on X that the company is still advancing robotics research and currently has 15 related positions open for recruitment.

However, compared with OpenAI's overall recruitment demand of more than 800 positions, this number is not large.

Some netizens pressed in the comments about whether OpenAI really plans to let Astra directly handle physical control. David Schnurr did not reply. Another netizen joked: "I like the way he doesn't reply."

II. Why No Longer Have a Separate Robotics Team?

Guanya Shi, an assistant professor at Carnegie Mellon University's Robotics Institute, believes that the signal sent by this adjustment is that OpenAI no longer regards robotics as a "special problem" that needs to be solved separately, but rather as a downstream application of general intelligence.

OpenAI's response is basically consistent with this judgment. The company said that at the beginning of this year, researchers still believed that the best way to advance robotics was to continue research and development on the basis of existing AI models such as Sora. To generate realistic videos, Sora needs to form a certain implicit understanding of the operating laws of the physical world.

And this capability is also generally believed to be related to robot control.

But since then, general frontier models, including GPT-6 Astra, have begun to show capabilities that can be directly applied to robots. It is precisely this change that prompted OpenAI to disband the independent team and move the relevant researchers into more general AI model research and development teams.

OpenAI also said that it reassigned robotics researchers to other teams because it now believes that this approach is more conducive to accelerating the realization of general embodied intelligence.

III. Beyond the Technical Route, There Is Also Cost Pressure

This reorganization also comes against the backdrop of OpenAI narrowing its business boundaries and controlling operating costs.

Last year, Sam Altman proposed that "side quests" must be abandoned. Subsequently, OpenAI cut some projects, including Sora, and refocused computing resources on more core businesses that are also more likely to generate revenue.

According to media reports, OpenAI's operating loss in the most recent quarter reached $10.9 billion (about 72.722 billion yuan).

Therefore, as it seeks to complete a new round of financing and prepares for a possible IPO in early 2027, how to control losses and focus on core products has become a question OpenAI must address.

Conclusion: General Models Enter Embodiment, Deployment Becomes the New Challenge

The core change in OpenAI's adjustment this time lies in the change in the way research and development is organized. Robotics is no longer advanced as an independent direction but is incorporated into the general model system.

Whether this judgment holds still needs time to verify. General models have shown the potential to directly control robots in simulated environments and some tasks, but moving from demonstrations to stable large-scale applications still involves practical issues such as reliability, safety, and engineering.

For the industry, the impact is already apparent. The narrative of betting on robot brains is being questioned, while companies with real scenarios, continuous deployment capabilities, and data accumulation may instead gain more favor.

OpenAI's consolidation of the team does not mean that the challenges of robot deployment are therefore over; it has merely entered a new stage. What is more important in this stage is to truly bring intelligence into the physical world.

Source

Zhidongxi