news

Huang Renxun talks with Zuckerberg: New chip samples will be sent this week, and the AI ​​industry still has 5 years of product innovation

2024-07-30

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Nvidia CEO Jen-Hsun Huang and Meta CEO Mark Zuckerberg rarely appeared on the same stage to "promote each other's business."

On July 29, local time, at the 50th SIGGRAPH graphics conference held in Denver, the United States, Huang Renxun and Zuckerberg had their first public fireside chat. In this relaxed conversation, the two talked about some of the company's latest progress and shared their views on AI trends. Zuckerberg said that basic AI research is "accelerating" and based on the current AI model technology, the industry still has five years of product innovation space.

Before the conversation began, Huang Renxun had already demonstrated a series of the company's latest products at the conference and announced that "Nvidia will start sending Blackwell samples this week, which is the company's first new chip architecture this year."


Huang Renxun and Zuckerberg had a dialogue on the same stage. Source: Nvidia Live

On the same day, Meta also announced the launch of a new tool called AI Studio, built on the company's latest big model Llama 3.1, allowing users to create, share and design personalized AI chatbots, while allowing Instagram creators to use AI characters "as an extension of themselves" to handle simple automatic replies.

Zuckerberg said in the conversation that as more and more information becomes available on social media, how to help users filter content has become a major problem, and generative AI can help improve the recommendation system of social media. He believes that in the future, every company will have its own AI agent that can independently complete complex tasks that take a long time.

Huang Renxun said that AI as a chatbot is still at the "one question, one answer" stage, and in the future, AI may be able to generate decision tree-like thinking for users: "When we are given a task or a question, you know we will consider multiple options, or conceive a decision tree and go down it, simulate it in our minds, and see the different results of each decision. AI may do the same thing in the future, which makes me super excited."

When Huang Renxun began to praise Meta's latest open source model Llama 3.1, which was released not long ago, Zuckerberg talked about the impact of the closed system of competitor Apple on him, which made him realize the importance of the open source ecosystem: "You not only need to build an (AI) software, you need an ecosystem around it. If we don't open source, it probably won't be so useful. We don't do this because we are altruists, although I think it will indeed help the ecosystem - we do this because open source will make what we are building the best."

Huang Renxun reiterated his optimism about the future development of humanoid robots during the conversation, believing that AI will have a model of the physical world. Zuckerberg talked about the company's long-term bet on smart glasses, saying that he had always thought that holographic AR glasses would appear before the AI ​​era, but the former has not been mass-produced until now; however, Zuckerberg still firmly believes that smart glasses will become the mobile phone version of the next generation computing platform.

Meta is one of Nvidia's major customers. In January this year, Zuckerberg announced that Meta was planning to build its own AGI (artificial general intelligence) and planned to obtain about 350,000 H100 GPUs from Nvidia by the end of this year. Including other GPUs, the total computing power owned by the company will be close to the computing power that 600,000 H100s can provide.

During the conversation, Huang Renxun also talked about some interesting things when he was with Zuckerberg. He said that Zuckerberg came to his house before, and they prepared cheese steak sandwiches that they both liked. When Zuckerberg was cutting tomatoes, Huang Renxun was surprised to find that the "obsessive-compulsive disorder" opponent cut each slice of tomato equally thin and placed them neatly in a circle on the plate, with the same distance between each slice: "That's why he needs an AI that won't judge him."

In addition, when talking about Nvidia's past days of "taking the long way around", Huang Renxun said with a smile that if he had known that it would take so many years to achieve success, in order to be one step ahead, he would have chosen to drop out of college like Zuckerberg.

After the conversation, Zuckerberg took out the gift he had prepared for Huang: a thick black leather jacket. Then the two exchanged jackets again.


Two people wearing leather jackets. Source: Nvidia Live

At this conference, NVIDIA announced a series of software updates. The company officially released NVIDIA NIM microservices for AI models, which can generate OpenUSD language to answer user queries, generate OpenUSD Python code, apply materials to 3D objects, understand 3D space and physics to help speed up the development of digital twins, etc.

NVIDIA also announced that it will provide a set of services, models and computing platforms to the world's leading robot manufacturers, AI model developers and software manufacturers to develop, train and build the next generation of humanoid robots. The first batch of companies to join NVIDIA's humanoid robot developer program include 1x, Boston Dynamics, ByteDance Research, FieldAl, Figure, Fourier, Galaxy General, Zhuji Power and dozens of other companies.

On the 29th, Nvidia (Nasdaq: NVDA) stock price closed at US$111.59 per share, down 1.30%, with a total market value of US$2.75 trillion; Meta (Nasdaq: META) stock price closed at US$465.71 per share, and the stock price closed flat, with a total market value of US$1.18 trillion.