OpenAI Chief Scientist Calls AI Progress Slow, Eyes Big Gains
OpenAI leaders anticipate faster AI advances in the months ahead. Chief scientist Jakub Pachocki shared this view with reporters at the GPT-5.5 launch. He predicted "pretty significant improvements in the short term, extremely significant improvements in the medium term." Pachocki pointed out that AI progress lately has felt "surprisingly slow."
OpenAI, founded in 2015 as a nonprofit research lab before shifting to a capped-profit structure in 2019, has driven many key developments in generative AI. The organization released landmark models like GPT-3 in 2020 and GPT-4 in 2023. Pachocki, who joined OpenAI in 2017 and became chief scientist after Ilya Sutskever's departure in 2024, oversees core research efforts.
Brockman's Take on GPT-5.5 Capabilities
OpenAI President Greg Brockman, a co-founder since the company's start, discussed GPT-5.5 in a podcast. He described it as a "new class of intelligence." The model stands out in programming tasks, creating presentations and spreadsheets, and operating web browsers. Brockman noted it concludes a two-year research push. Yet he added, "in many ways, it's a beginning point."
Brockman has long championed OpenAI's scaling approach to AI. Under his leadership alongside CEO Sam Altman, the company raised billions in funding from investors like Microsoft. GPT-5.5 builds on prior releases, pushing boundaries in practical applications.
Foundation for Next Reasoning Models
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
GPT-5.5 may lay groundwork for advanced reasoning systems. It could mirror how GPT-4o supported the o-series models: o1, o3, and o4-mini. Those models address tough problems by adding compute during inference. This method boosts performance on complex reasoning without solely relying on larger training runs.
OpenAI introduced the o1 model in late 2024 as a reasoning-focused system. It previewed chains of thought to solve math and coding challenges better. Subsequent versions like o3 and o4-mini refined this for speed and cost. GPT-5.5's stronger base might enable even more capable successors.
Doubts from AI Researchers
Not all experts share OpenAI's optimism. More AI researchers question the path of scaling language models, a strategy central to labs like OpenAI, Anthropic, and Google DeepMind. They view it as a potential dead end for true intelligence. Instead, some advocate fresh architectures to reach general AI.
This debate echoes ongoing tensions in AI. Critics argue pure prediction from vast data hits limits. Proponents of new paradigms explore neurosymbolic methods or hybrid systems. OpenAI sticks to its compute-heavy roadmap for now.
OpenAI continues to shape the field with releases like GPT-5.5. Expectations run high for the promised gains amid mixed views on the trajectory.