Meta has been open sourcing almost all of its AI creations in recent years, and the company has been praised for its commitment to open source. A new report however raised concerns about Meta’s plans to open source a GPT-5 level model.
The report, which was posted on the X forum (formerly known as Twitter) by Jason Wei from OpenAI, claims that Meta plans to build Llama 3 and 4, which are expected to be much more powerful than GPT-4. When asked if the company would still open source these models, a Meta AI employee reportedly said that they would, but that they were “sorry, alignment people.”
The term “alignment” refers to the idea of ensuring that AI systems are aligned with human values. There are concerns that open sourcing a GPT-5 level model could make it easier for bad actors to use the model for harmful purposes.
In a recent interview, Sam Altman, the CEO of OpenAI, said that the company has a “kill switch” that can be used to shut down its AI systems if they get out of control. However, it is unclear if Meta has any similar safeguards in place.
Meta has not yet commented on the report.
The sources for this piece include an article in AnalyticsIndiaMag.