Xiaoyong Shen: Don't Talk About Large Models Without 10,000 Cards!

Xiaoyong Shen: Don't Talk About Large Models Without 10,000 Cards!On September 5th, at the 2024 InclusionBund Summit, Xiaoyong Shen, a foreign academician of the US National Academy of Engineering and the UK Royal Academy of Engineering, made a blunt statement, saying that a large language model company shouldn't even claim to be developing large models without 10,000 cards. Shen humorously remarked, Talking about cards hurts feelings, but no cards mean no feelings

Xiaoyong Shen: Don't Talk About Large Models Without 10,000 Cards!

On September 5th, at the 2024 InclusionBund Summit, Xiaoyong Shen, a foreign academician of the US National Academy of Engineering and the UK Royal Academy of Engineering, made a blunt statement, saying that a large language model company shouldn't even claim to be developing large models without 10,000 cards. Shen humorously remarked, Talking about cards hurts feelings, but no cards mean no feelings. He emphasized that companies need to attract talent by providing sufficient resources, as research often requires 100 or even 1000 cards. If companies cannot provide these resources, talent will naturally choose not to join.

Shen's remarks directly address the resource bottleneck currently faced by large model development. In recent years, with the development of large model technology, the demand for computing power has exploded, and the scarcity of computing resources has become a key factor hindering progress. Regarding the next step in large model development, Shen believes that it will inevitably move towards embodied intelligence, with autonomous driving being a prime example of its application.

Embodied intelligence refers to the close integration of artificial intelligence with the physical world, allowing AI to perceive, understand, and manipulate the real world. Autonomous vehicles are a prime example of this integration, combining artificial intelligence technology with vehicle hardware to achieve autonomous driving capabilities. Shen's view also reflects that the development of large models goes beyond just algorithms and data; it requires close integration with the real world to truly unlock its potential.

Shen's remarks have sparked heated discussions within the industry, highlighting the need for companies to increase investment and provide sufficient resources to talent in the face of fierce competition in the field of large models. This is crucial for driving further breakthroughs in large model technology.