Meta AI Shows First Signs of Self-Improvement, Zuckerberg Warns of Tighter Release Controls
Meta has reported that its artificial intelligence systems are beginning to improve themselves without human intervention, a development CEO Mark Zuckerberg says marks «the first step towards achieving artificial superintelligence (ASI).»
«Over the last few months we have begun to see glimpses of our AI systems improving themselves,» Zuckerberg wrote in a policy paper on July 30. «The improvement is slow for now, but undeniable.»
ASI refers to AI that can far exceed human intelligence and self-enhance at an exponential rate a concept long debated in the tech world. While optimistic about its potential, Zuckerberg stressed that Meta will no longer openly release its most powerful AI systems.
«I am extremely optimistic that superintelligence will help humanity accelerate our pace of progress,» he said. «But perhaps even more important is that superintelligence has the potential to begin a new era of personal empowerment.»
The Meta chief argued that personal AI could help people «achieve your goals, create what you want to see in the world… and grow to become the person you aspire to be.»
His comments follow similar research into AI self-improvement, including a 2024 University of California, Santa Barbara study on a «Gödel Agent» capable of rewriting its own code and proving the benefits of each change.
Zuckerberg said ASI could lead to «the discovery of new things that aren’t imaginable today» — but the path to that future will be more tightly controlled than before.
Kursiv Uzbekistan also reports that Microsoft relentless investment in datacenters and AI talent powers record growth and stock surge.
