OpenAI says its latest creation may be the moment artificial intelligence crosses a significant line, Axios reports. On Thursday it rolled out GPT-6 Astra, which president Greg Brockman described as a "generational leap" and suggested could mark the arrival of artificial general intelligence, telling reporters, "Welcome to the AGI era." As VentureBeat explains, OpenAI defines AGI as "highly autonomous systems that outperform humans at most economically valuable work."
Astra was trained using more than 100,000 GPUs and is OpenAI's first model deemed "critical" under its own cybersecurity risk scale, capable of autonomously hunting for serious software flaws. It can plug directly into software to draft legal contracts, design circuit boards, build 3D games, and help with advanced scientific work, according to OpenAI demos. The model will debut to select organizations before reaching paying users, as the company confronts mounting concerns over safety, monitoring, and whether Astra is becoming harder for humans to oversee. In August, OpenAI said it would be pausing training of new models due to cybersecurity concerns, CNET reports.