Mark Zuckerberg says Meta “should be careful about what we chose to open source,” citing new risks from super intelligence.
Mark Zuckerberg has published his AI manifesto, claiming the type of “personal tension” that people can use to achieve their individual goals.
in New blog post, The CEO of Meta said, “We want to build personalized AI that will help you achieve your goals, create what you want to see in the world, become better friends, and grow to be the person you want. But the company’s new purpose comes with warnings. This powerful AI may be too strong to be open to the world any time soon.
“We believe that the benefits of superintelligence should be shared as widely as possible with the world. It says that superintelligence raises new safety concerns,” writes Zuckerberg. “We need to be strict about reducing these risks and be aware of those who choose to be open source. Still, we believe that building a free society requires us to aim to empower people as much as possible.”
Among those risks, he wrote that AI could become “a force focused on replacing the big belts of society.”
Zuckerberg has traditionally placed meta as a supporter of open source AI, especially compared to rivals such as Openai and Openai. Google. While many argue that the company’s llama model does not meet the strict definition of “open source,” the company is more leaning towards open-sourcing of frontier models than most major high-tech peers.
in Last year’s blog post, Zuckerberg created a passionate case of open source, telling Meta that it is “the next step to becoming an industry standard.”
“I think open source is necessary for a future of positive AI,” Zuckerberg wrote last year. “Open source will allow more people around the world to access AI benefits and opportunities, and its power is not concentrated in the hands of a few companies, allowing technology to be deployed more equally and safely across society.”
The CEO left himself with some wiggling rooms. Podcasts If there was a major change in AI capabilities last year, it may not be safe to “open source” them.
The closed model gives companies more control over product monetization. Zuckerberg pointed out last year that “Llama’s release does not undermine our ability to invest in revenue, sustainability, or research like a closure provider.” In contrast to competitors like Openai, Meta makes most of the money from selling internet advertising.
Closed vs. Open Source AI
AI safety experts have long debated whether open source models are responsible for advanced AI development. Some argue that the open-sourcing AI model will allow for broader scrutiny to democratize access, accelerate innovation and improve safety and reliability. However, some say that publicly releasing powerful AI models can increase the risk of misuse by bad actors, including misinformation, cyberattacks, and biological threats.
There is also commercial debate about open source. That’s why most major AI labs keep their models private. A powerful AI model of open-sourcing can erode a company’s competitiveness by allowing rivals to copy, tweak or commoditize core technologies.
As Zuckerberg said last year that Meta’s business doesn’t rely on selling access to AI models, Meta is in a different position from its rivals here. “Releasing Llama does not undermine your ability to invest in revenue, sustainability, or research, like a closed provider,” he said.
Meta representatives did not immediately respond to requests for comment from luck, Made outside of normal working hours.