Why do companies like X Square deploy imperfect AI robots in real-world settings?

Question

Accepted Answer

Deploying Imperfect AI Robots in Real-World Settings Facilitates Crucial Data Collection for Embodied AI Development

Companies such as X Square deploy imperfect AI robots in real-world environments primarily as a method for gathering essential data. This approach allows for the collection of real-world datasets, which are indispensable for training embodied artificial intelligence. Unlike large language models that learn from vast amounts of internet content, embodied AI requires interaction with the physical world to develop. Studying the robots' performance in unfamiliar settings provides invaluable data that drives the robot's growth and enhances its capabilities, proving more informative than relying solely on controlled laboratory experiments.
Gathering Real-World Data for Embodied AI

The deployment of AI robots in actual domestic settings, such as the cleaning service offered by 58.com and X Square, serves as a crucial data-gathering exercise. This strategy is specifically designed to collect real-world datasets, a necessity for advancing embodied artificial intelligence. This contrasts with the training methodologies for large language models, which are trained on extensive internet-based content. The performance data generated from these real-world applications is vital for the continuous improvement and development of the robots.
Enhancing Robot Development Through Unfamiliar Environments

By placing robots in completely unfamiliar environments, companies like X Square gather data that is highly beneficial for the robot's development. This approach presents significant challenges but yields unique data that aids in the robot's learning process. This method is considered more informative than confining the robots to laboratory settings, as it exposes them to a wider range of unpredictable scenarios and interactions.
Limitations and Future Potential of Real-World AI Applications

While the deployment of these robots offers valuable data, their current capabilities are limited, as demonstrated by the Quanta X1 Pro's rudimentary clothes-folding process, which resembles a child learning the task. This highlights the ongoing challenge in matching human dexterity. For instance, while robots may be able to perform tasks like picking up rubbish, their ability to fold clothes still requires significant development, taking several minutes for a single item.
Significant Investment in Embodied AI Industry

The broader embodied AI industry in China is experiencing substantial investment, with investors pouring over 57.7 billion yuan (US$8.5 billion) into the sector this year alone. This figure already surpasses the total investment for the entirety of last year. This surge in funding underscores the growing interest and potential perceived in embodied AI technologies, leading to similar trials such as robots directing traffic in Hangzhou and operating on factory floors.

Reading history

Hide →

This history is cleared when the page is closed.

Why do companies like X Square deploy imperfect AI robots in real-world settings?

Loading

Was this article helpful?

Helpful

Not helpful

Report an issue

1

0

0

Share with friends