on september 15th, in an interview, a16zOpenAI Co-founder Greg BrockmannGreg BrockmanThe blogger says:Humankind has entered the age of universal artificial intelligence。
1AI Note: Artificial General IntelligenceAGIUSUALLY REFERS TO AI, WHICH CAN HANDLE COMPLEX COGNITIVE TASKS IN A WIDE RANGE OF AREAS AND CAN MIGRATE, LEARN AND COMPLETE NEW TASKS。
OpenAI, in OpenAI Charter, defines AGI as “a system that is highly autonomous and transcends human beings in most economically valuable jobs”. This definition contains two thresholds: first, that the system can autonomously complete a more complete work chain; and second, that in most economic activities, its performance exceeds that of humans。
During the interview, Brockmann indicated that AI had reached the threshold that could be called the “AGI era” with models such as GPT-6 Astra having long-term autonomous implementation and computer-operated capabilities. AGI is no longer just a concept of the future, but is in realityTHE NEXT KEY ISSUE WILL GRADUALLY MOVE FROM “WHEN TO ACHIEVE” TO HOW TO USE AND DEPLOY INCREASINGLY POWERFUL AI SYSTEMS。
In response, Brokman indicated that OpenAI had redeployed 25% production engineers, suspended their original projects and moved to system defence and security infrastructure upgrades. The team used Astra to identify gaps in the internal system and stated that a number of serious problems had been repaired。
In addition, Brockman stated that the team had discovered new problems during the check-up, but subsequently they had been found to be “saturated”. He stated that, to the knowledge of OpenAI, the team had largely found all the P0 issues that Astra was able to identify and all the key issues that it could find。
He noted that that result did not mean that there were no other gaps in the system. With the emergence of the new model, the team will initiate a new round of rehearsing and deploy a new generation of network security capabilities to the internal system to find previously unidentified loopholes。