The Gong Dou incident made the market aware of the instability of exclusive models. Many companies will tend to deploy open source models to hedge risks. Meta may become the giant that silently wins. The farce of the court battle at OpenAI, the world's most famous technology startup, ended with CEO Altman returning to the helm within 100 hours. But the undercurrent of competition among major technology companies is not over yet.

Some analysts pointed out that compared to Microsoft, which turned the tide in the crisis and desperately avoided letting its tens of billions of dollars of investment go to waste, Meta, which was not really involved in the palace dispute, may be the giant that silently wins. The reason is that this palace farce has made the market realize that exclusive models like OpenAI's GPT large model are too closely tied to one company and have too much instability.

The battle between open source and closed routes

In the AI ​​wave, there are two opposing factions. One is the "closed-source faction" represented by companies such as OpenAI and Anthropic. The technical details of these companies' large models are not disclosed to the public and the code is under their own control; the other is the "open-source faction" represented by Meta. The code is open to the public and everyone can modify or even re-develop the code within the scope of copyright restrictions.

People who support the "open source faction" believe that open source is conducive to the rapid establishment and development of the ecosystem, and can also gather global forces to help AI models quickly iterate. Open source is to unlock the wisdom of the group. Decentralization and combating excessive concentration of power.

But those who support the "closed source" point out that open source is harmful to creators, competitors can copy their works, and it may be potentially dangerous, and someone may use the most advanced AI model to do bad things.

OpenAI chief scientist Ilya Sutskever once made it clear:

"If you believe AGI is coming, open source is a bad idea."

Mark Surman, developer of the Firefox browser and president and executive director of the Mozilla Foundation, an Internet open source organization, told the media:

“The OpenAI/Microsoft drama highlights one of the near-term risks of artificial intelligence, that the next wave of technology will be controlled by the same small group of players that shaped the last Internet era... If GPT-X can be responsibly open sourced and give researchers and startups the opportunity to make this technology safer, more useful, and more trustworthy, then we may have a chance to avoid this situation.”

Just a few weeks ago, Mozilla teamed up with Meta chief artificial intelligence scientist Yann LeCun and other celebrities to publish an open letter calling for the AI ​​industry to be more open. The open letter has received more than 1,700 signatures.

The open letter reads:

"Increased public access and scrutiny will make technology safer, not more dangerous. The idea that tight proprietary controls on underlying AI models are the only way to protect us from large-scale societal harm is naive at best and dangerous at worst."

The palace fight may affect OpenAI’s business

Over the weekend, hundreds of OpenAI customers reportedly began contacting OpenAI's competitors, including Anthropic, Google, and Cohere, worried that their businesses might be affected if OpenAI disintegrated overnight.

Their concerns are not without reason. The cloud computing industry is a typical example. Luis Ceze, professor of computer science at the University of Washington and CEO of OctoML, told the media:

"There's a danger in putting all your chips in one basket, and we saw this in the early days of cloud computing, which led to companies moving to multi-vendor and hybrid environments."

Compared with OpenAI's GPT large model, the capability gap of Meta's open source model is not large, so it is also very cost-effective. It is an ideal channel for companies to achieve diversified AI configurations. Enterprises that rely on proprietary models can quickly minimize business risks by deploying open source models.

Some analysts said that OpenAI's short-lived palace farce highlighted the "grass-roots" and imperfect governance of small companies. Some technical staff who care about stability may be more willing to join Meta's AI research laboratory.