Li Zhifei: Elon Musk's xAI is cautiously optimistic

Source: Fei Ge said AI

Author: Li Zhifei

Image source: Generated by Unbounded AI‌

Another “burst” was seen in the morning: Elon Musk’s xAI team was established. It’s okay to be free, Brother Fei, comment:

1. Star team: The team has assembled members from major companies such as OpenAI, Deepmind, Google, etc., which is very strong. But there is always a feeling that it is necessary to have all kinds of ingredients in the vegetable market, fish, meat and vegetable soup. Can a “get together” team really last long?

2. Money: The current mainstream of refining large models is violent aesthetics and vigorously producing miracles. If the money is not enough, you can only find another way outside the mainstream, but the results will be slower. As the richest man in the world, Musk’s ability to make money is also invincible in the universe. It is estimated that a lot of local tyrants are waiting in line to give money, so money should not be a problem in the early stage, but there will be a lot of uncertainty in the later stage.

3. Direction: Elon has his own scene, Twitter and Tesla, so it can be more targeted when refining the big model. But Twitter is a bit small, and the large model for autonomous driving is still completely uninhabited, so it cannot be said to be an advantage (relative to strong competitors). As for the insecurity and inauthenticity of focusing on the existing large models, I personally feel that they have no unique tricks. Safety and authenticity are not the main dishes of large models at present, but the ability of models is, so it can only be regarded as the direction of PR.

4. Follow-up organizational culture: Although building a large-scale model requires top down to determine the strategic direction, the implementation of AI research and development still requires a more bottom-up culture. But judging from the history of Musk’s governance of Twitter, it should be difficult to establish a typical Silicon Valley Internet company culture. I know that many Silicon Valley engineers are disgusted with Musk’s tyrant style. Therefore, this is a negative score, which will not be good for subsequent recruitment and retention.

5. Innovation: I feel that this is a typical accumulation situation. Recently, there is a popular saying in the circle of friends that “greatness cannot be planned”.

6. Business model: no

To sum up, although Musk is awesome, it is hard to say whether he can distort the field of reality and violate the first principles he advocates to achieve great success.

Of course, spending money on AI, promoting a group of engineers, acting as AI to compete with squid, and having a new topic on Twitter to entertain the masses are all worthy of encouragement and expectation.