Business News | AI Robotics May Need 10 Million Hours of Interaction Data to Reach Production-level Capabilities: Goldman Sachs

New Delhi [India], October 30 (ANI): General-purpose artificial intelligence (AI) robotics could require an estimated 10 million hours of high-quality physical interaction data to reach production-level capabilities, highlighting the challenges in achieving commercial viability, even as technological progress in the sector is projected to be two to four times faster than the historical pace of autonomous driving technology, according to a report by Goldman Sachs.

The report, titled "China AI Robotics: Expert Call on Frontier Foundation Models in Physical AI & Embodied Robotics," highlighted that advances in frontier AI models are accelerating the development of intelligent robots, but limitations in physical interaction data, precision and processing speed remain major barriers to widespread commercialisation.

Also Read | Jiyaa Shankar Unfollows Fiance Kaaran Dhanak on Instagram Amid Cryptic Posts, Sparks Breakup Rumours.

Drawing on an October 28 interaction with a robotics algorithm engineer, the report estimated that embodied AI would require approximately 10 million hours of high-quality, multimodal interaction data to achieve generalised production capabilities.

"Embodied AI will likely require an estimated 10 million hours of high-quality, multi-modal interaction data," the report noted, emphasising the importance of reliable physical interaction data rather than passive or noisy recordings.

Also Read | Google Headcount Grows by 10,000 Amid Tech Industry Layoffs.

The substantial data requirement suggests that achieving commercial viability for general-purpose robots will depend not only on advances in AI models but also on the industry's ability to collect high-quality information about real-world physical interactions.

Unlike autonomous vehicles, which primarily operate through steering, acceleration and braking, advanced robotic systems must handle dozens of mechanical degrees of freedom, continuous physical contact, shear forces and tactile feedback.

The report noted that general-purpose robotic manipulation remains at an early stage of maturity, comparable to autonomous driving technology before 2015 or around 2017–2018, when assessed against safety-critical, production-grade reliability requirements.

However, it projected that technological progress in physical AI could be two to four times faster than the historical pace of autonomous driving development, supported by the growing ability of foundation models to understand unfamiliar environments and perform new tasks without extensive additional training.

The report identified three critical engineering bottlenecks: three-dimensional (3D) spatial perception, inference latency and inference economics.

On spatial perception, the report noted that "current spatial fidelity remains insufficient for millimeter-level physical interactions," resulting in repeated trial-and-error during robotic operations.

Similarly, current processing speeds remain inadequate for real-time dynamic stabilisation, high-speed manipulation and microsecond-level reflexive recovery.

On inference economics, the report highlighted the need to transfer advanced physical reasoning and spatial awareness capabilities from large AI models into smaller, computationally efficient models capable of operating directly on robotic hardware.

The report further noted that industrial manufacturing requires six-sigma reliability and microsecond-level repeatability, while general-purpose AI models are better suited to less structured environments where adaptability is more important than sub-millimetre precision.

This indicates that improvements in AI reasoning alone may not be sufficient to make general-purpose robotics commercially viable without corresponding advances in physical precision, reliability and real-time execution.

The report also highlighted limitations in existing data-collection methods.

It noted that first-person video recordings can help AI models understand tasks and user intentions but lack information about physical forces and contact dynamics.

Similarly, physics-based simulations can accelerate training but struggle to accurately reproduce real-world friction, material deformation and physical interactions.

The report identified dexterous Universal Manipulation Interface (UMI) systems, supported by high-resolution tactile sensing, as a promising approach to collecting high-quality physical interaction data.

It suggested that combining direct touch information with advanced sensing technologies could help address the precision limitations of vision-based systems.

Goldman Sachs also identified advanced 3D spatial sensors, high-resolution tactile arrays and physical data-collection infrastructure as potential near-term beneficiaries of increasing demand for embodied AI.

On the industry's commercial outlook, the report highlighted "Shifting moats toward edge and hardware integration," suggesting that robotics manufacturers will increasingly differentiate themselves through hardware integration, real-time motion control, computational efficiency and the ability to deploy AI models directly on robotic devices.

It further suggested that collaborative robots and existing industrial automation systems could offer more practical near-term commercial opportunities than humanoid robots alone, particularly in manufacturing environments requiring high precision and reliability.

While frontier AI models are expected to accelerate technological development, the report characterised the industry's trajectory as "An acceleration rather than an overnight inflection," indicating that technological improvements may bring forward commercial deployment without immediately resolving the challenges facing general-purpose robotics. (ANI)

(The above story is verified and authored by ANI staff, ANI is South Asia's leading multimedia news agency with over 100 bureaus in India, South Asia and across the globe. ANI brings the latest news on Politics and Current Affairs in India & around the World, Sports, Health, Fitness, Entertainment, & News. The views appearing in the above post do not reflect the opinions of LatestLY)