share_log

Meta AI高管暗批OpenAI:闭源AI是荒谬的 开源是唯一出路

Meta AI executives criticize OpenAI, stating that closed-source AI is absurd and open-source is the only way forward.

Sina Technology ·  Dec 11 03:58
big

On the evening of December 11, Beijing time, Manohar Paluri, Vice President of Generative AI Engineering at Meta, recently stated that closed-source AI is absurd, and to fully realize the potential of AI, open source is the only way forward.

The prosperity of AI has once again ignited a long-standing debate in the Technology Industry regarding the merits of open source Software versus closed source Software. However, for Meta, the parent company of Facebook, providing open-source large language models (LLMs) is the only way to fully harness AI's potential.

Paluri stated at the Fortune AI Brainstorming Conference held in San Francisco: "I think a world providing closed-source AI accessible only through specific APIs is absurd. Only by walking the open path can AI make progress in the coming decades."

As is well known, the pioneer of generative AI, OpenAI, currently adopts a closed-source model, with its main revenue sources including subscriptions and API services. As a co-founder of OpenAI, Tesla CEO Elon Musk sued OpenAI because the company betrayed its original mission by abandoning its open-source commitment.

Musk is concerned that if OpenAI continues down this path, it will limit the development of AI technology, and issues of safety and ethics will not receive adequate attention, which could ultimately have adverse effects on society as a whole. Therefore, Musk urges OpenAI to reassess its commitment to open source and take action to make its products publicly accessible again.

From a technical standpoint, Meta's Llama large language model does not currently meet the definition of open source as updated by the Open Source Initiative (OSI) a month ago, yet Llama remains one of the most powerful open-source AI models available for the public to download, customize, and create AI applications or tools.

Paluri believes that the accessibility of Llama (i.e., open source) has been key to its success, as "distribution" is one of four elements vital to the success of any AI model today, alongside talent, computing power, and data.

Critics of open-source AI mainly express concerns about the technology falling into the hands of bad actors. When asked about security issues, Paluri did not answer directly but emphasized the benefits of open-source.

With its accessibility, Llama has helped Meta catch up with competitors OpenAI and Alphabet in the AI race. Initially, at the end of 2022, the emergence of generative AI caught Meta off guard.

In recent weeks, Meta has repeatedly stated that its Meta AI (generative AI chatbot and search bar) has approximately 0.6 billion active users per month. In comparison, OpenAI CEO Sam Altman mentioned earlier this month that ChatGPT, which operates under a closed-source model, has 0.3 billion users weekly. OpenAI aims to increase the active user count to 1 billion within the next year.

It is currently unclear how many of the 0.6 billion monthly users reported by Meta AI are actively seeking to interact with the AI, rather than just using services like search that use AI behind the scenes. When asked about the users, Paluri declined to provide more detailed segmentation. He pointed out that it is difficult to explain users' specific intentions when accessing a feature and mentioned that the technology works best when it is 'seamless.'

Although ChatGPT was the first to spark the wave of generative AI chatbots and was over a year ahead of Meta AI, Paluri stated that Meta AI's popularity is also driven by the usage of nearly all Meta applications, including WhatsApp, Instagram, and Facebook, which collectively have over 3 billion users daily.

Paluri said, 'At the beginning of this year, Meta AI was just a starting point of what we were doing. Now, this technology is improving day by day.'

As Meta's CEO, Mark Zuckerberg has also been advocating for open-source AI, believing it is the direction of future development. Zuckerberg believes that although some AI companies are developing leading closed-source models, open-source AI is rapidly closing the gap, and the open-source model is more conducive to company growth.

Zuckerberg emphasized that open-source can promote the development of a complete ecosystem and keep various models in the industry competitive, efficient, and open. From a longer-term perspective, open-source will make AI safer and more transparent.

Disclaimer: This content is for informational and educational purposes only and does not constitute a recommendation or endorsement of any specific investment or investment strategy. Read more
    Write a comment