OpenAI首席科学家称AI进展“出人意料地慢”，并承诺未来将有重大飞跃

OpenAI's leadership expects AI development to pick up speed in the coming months. At the release of GPT-5.5, chief scientist Jakub Pachocki told reporters to expect "pretty significant improvements in the short term, extremely significant improvements in the medium term," arguing that recent AI progress has been "surprisingly slow."

In a podcast, OpenAI President Greg Brockman called GPT-5.5 a "new class of intelligence" that excels at programming, building presentations and spreadsheets, and using browsers. The model caps off a two-year research effort, Brockman says, but "in many ways, it's a beginning point."

The likely reasoning behind the acceleration thesis: GPT-5.5 could serve as the foundation for a new generation of more efficient reasoning models, just as GPT-4o underpinned the o-series—o1, o3, and o4-mini—which tackle complex tasks through additional inference-time compute. Not everyone agrees, though. A growing number of AI researchers see the language model approach favored by the big labs as a dead end and are pushing for new architectures to achieve real intelligence.