Mark Zuckerberg says that dead should be “careful about what we choose to open the source,” noting new risks from Super Intelligence

in A new blog post, Meta CEO said that he wants to build a designated Amnesty International that helps you “achieve your goals, create what you want to see in the world, to be a better friend, and grow to become the person you aspire to be.” However, the new goals of the company come with a warning: this strong artificial intelligence may be very strong so that it cannot be left open to the world.

“We believe that the benefits of adhesion with the world should be shared widely as possible. However, Superincyligence will raise fears related to new safety,” Zuckerberg wrote. “We will need to be strict about mitigating these risks and warning about what we choose to open the source. However, we believe that building a free society requires aiming to enable people as much as possible.”

Among these risks: Artificial intelligence can become “a force that focuses on replacing large areas of society.”

Zuckerberg traditionally placed in a dead position as an open source Amnesty International, especially compared to competitors such as Openai and Google. While many argue that the company’s Llama models do not meet the strict definition of the “open source”, the company has relied on more open models of sources in most of its peers in technology.

in Blog post last year, Zuckerberg has provided an enthusiastic issue for the open source, as she preaches dead as “the next steps towards Open Source AI are the industrial standard.”

“I think the open source is necessary for the positive future of Amnesty International,” Zuckerberg wrote last year. “The open source will guarantee that more people around the world can reach the benefits and opportunities of artificial intelligence, and that this force is not concentrated in the hands of a small number of companies, and that technology can be spread evenly and safely through society.”

The CEO himself left a space for maneuvering, saying in A. Podcast Last year, if there is a major change in the capabilities of artificial intelligence, it may not be safe “open source”.

Closed Models give companies more control over their products. Zuckerberg pointed out last year that Meta does not depend on selling access to artificial intelligence models, so “Llama’s launch does not undermine our revenues, our sustainability or our ability to invest in research as does it for closed service providers.” Unlike competitors like Openai, Meta makes most of its money selling ads online.

Closed against the open source AI

Artificial intelligence safety experts have long discussed whether they were open or closed models more responsible for developing advanced artificial intelligence. Some argue that artificial intelligence models are open source destroying, accelerating innovation, and allowing the broader scrutiny to improve safety and reliability. But others say that the launch of powerful international organization models may publicly increase the risk of misuse by bad actors, including wrong information, electronic attacks or biological threats.

There is a commercial argument against Open Source, which is why most of the artificial intelligence laboratories keep their models in particular. Open source artificial intelligence models can be eroded by allowing competitors to copy, adjust or assemble their basic technology.

Meta is in a different position here from some of its competitors, as Zuckerberg said last year that Meta does not depend on selling access to artificial intelligence models. He said: “The launch of Lama does not undermine our revenues, our sustainability, or our ability to invest in research as it does for the closed service providers.”

Meta representatives did not immediately respond to a request for comment from luck, Made outside normal working hours.