.png){kind=logo}
Mark Zuckerberg, CEO of Meta, recently shared his new perspective on the practice of open-sourcing AI models. This marks a significant shift from his previous stance, where he openly dismissed closed platforms with a strong declaration of “f*** that.” This change comes as Meta focuses on its long-term goal of developing Superintelligence.
Mark expressed concerns about the potential for misuse and safety issues that arise as the technology continues to advance. During Meta’s second-quarter earnings call, he mentioned that there could be policy shifts regarding the current approach to open-source platforms.
Zuckerberg expressed his concerns with open-sourcing AI models, saying there has always been a conflict between being helpful and productive. He believes that with the new AI models being highly capable, there is room for misuse and harm, so it is more reliable to shift to closed platforms.
He clarified his position, saying Meta has always open-sourced some of its models but not everything they’ve built, and will continue to produce and share leading open-source models. He further added, “Obviously, as you approach real Superintelligence, I think there's a whole different set of safety concerns that I think we need to take very seriously.”
Zuckerberg pointed out the impracticality of open-sourcing large AI models since they are not widely used. He said this when talking about his vision of building ‘Superintelligence.’ He further added that these generative AI models can be exposed to malicious activities and be repurposed to fulfill malevolent intentions.
Mark’s fears are not baseless, as these smart machines can affect the outcomes of elections and increase the possibility of cyberattacks.
Also Read: Mark Zuckerberg Announces Next-Gen Ray-Ban Smart Glasses for 2025
Open-sourcing AI models has always been a significant concern for developers in the field. Mark Zuckerberg, who previously supported the idea of open-sourcing AI platforms, has recently changed his stance. His vision of creating superintelligence has led him to reconsider this approach.
