American Foreign Member of the National Academy of Engineering, Shen Xiangyang: Large models will dominate all vertical industries
ZhangJing
Tue, Mar 26 2024 06:26 AM EST
- From the universe to quantum mechanics, physics has a unified theory called the Theory of Everything. Today, deep learning does not have such a theory, so it is called the Theory of Anything.
- Computing power is a threshold. "To have a powerful model requires a lot of resources. There used to be a saying that poverty limits imagination. Now, poverty may distort imagination because without sufficient resources, the projects we can imagine might be different."
On March 23, 2024, the Global Developer Pioneer Conference opened in Shanghai. Shen Xiangyang, a foreign member of the National Academy of Engineering in the United States, shared his thoughts on large models in his speech titled "Opportunities and Challenges in the Era of Large Models". He stated that artificial intelligence is flourishing, but people do not have a clear understanding of the nature of intelligence. Physics has the Theory of Everything, and deep learning is the Theory of Anything.
In the past few years, the development of large models has been booming, and the open-source community has shown great vitality and vitality. Shen Xiangyang said that Words and Pictures enable a thousand words to be conveyed with a single picture, and ChatGPT enables one sentence to be equivalent to ten thousand sentences. The emergence of Sora enables the generation of impressive results with just a few prompts.
These three things have caused anxiety about where the opportunities for future developers lie. Shen Xiangyang believes that multimodality is the path to general artificial intelligence. The technologies represented by OpenAI actually involve understanding different modalities. "OpenAI has two internal systems: one extends from Words and Pictures to Words and Videos, fol...
