Earlier today, Mark Zuckerberg shared a rembing memo, pointing to his vision to make the AI “Superintendentus”. In the memo, Zuckerberg indicated that the company may need to be more selective in its open sources to achieve a more powerful AI.
Referring to “safety concerns”, he wrote that Meta will need to be “tough” about such decisions. The line was standing in front of many people like Zuckerberg – who once called “Fuck” in relation to closed platforms – made the open source the center of Meta’s AI strategy.
During the second quarter income call of Meta, Zuckerberg further acknowledged that there could be a change, though it reduced its importance. When he was asked if his thinking had changed, what did he say here?
I don’t think our thinking has changed especially on it. We have always been open to some of our models and not open what we have done. So I would expect that we will continue to develop and share the open source model. I also think that there are many trends that are playing. One is that we are getting models that are so big that they are not practically practical to use many other people, so we do such a boat whether it is fruitful or helpful in sharing it, or if it is really helping rivals or something like that. So I think it’s there, he is concerned.
And then, obviously, when you approach the real sprintillence, I think there is a very different combination of safety concerns that I think we need to take very seriously, about which I wrote in my note this morning. But I think the most important thing is that I would expect that we will continue to work open source. I expect that we will be a leader there, and I also expect that whatever we do, we will not open the source, which is a kind of continuity in which we, what we are, are working in a way.
This is especially different from what he wrote about a year ago in a different memo, titled “Open Source AI is a way forward.” In this, even a long note, he said that open source is very important for both meta and developers.
He wrote, “People often ask if I am worried about leaving the technical advantage by open souring Lilama, but I think it loses a bigger picture.” “I expect the development of AI to be very competitive, which means that the open model of any model is not currently benefiting the next best models.”
He also argued that open source is Unharmed. “There is an ongoing debate about the safety of open source AI models, and I think open source will be more secure than AI alternatives. As long as everyone has access to similar breeders – which promotes open source – then more computing governments and institutions will be able to check bad actors with less computers.”
Clearly, Zuckerberg said the company will continue to work in some of its work. But it seems that it is laying the foundation for a future in which Meta’s “superstitles” can be rarely open.
