Meta says it's starting to see the first signs of AI systems that can improve themselves.
"Over the last few months we have begun to see glimpses of our AI systems improving themselves. The improvement is slow for now, but undeniable," CEO Mark Zuckerberg writes in a policy paper on the future of superintelligence.
This shift toward self-optimizing AI could mark a turning point. Some researchers believe it could dramatically speed up progress toward superintelligence and introduce new dynamics in how AI develops. "Developing superintelligence is now in sight," Zuckerberg writes.
At the same time, Zuckerberg is signaling a more cautious approach to open source. Unlike previous Llama models, it doesn't seem like Meta plans to release its most advanced AI to the public.
"Superintelligence will raise novel safety concerns. We'll need to be rigorous about mitigating these risks and careful about what we choose to open source. Still, we believe that building a free society requires that we aim to empower people as much as possible."
In the last few months, Meta has gone on a hiring spree, reportedly offering AI researchers from leading AI labs and Apple signing bonuses of up to several hundred million and, in some cases, even higher to join its newly formed superintelligence team—with some reports suggesting offers reaching more than a billion dollars. Meta CTO Andrew Bosworth confirmed these price regions, stating that its "complex compensation deals" are spread out over several years.
