According to OpenAI's CEO Sam Altman, the next generation of AI systems, represented by the new o1 model, will be capable of performing tasks autonomously without human input. During a discussion at T-Mobile's Capital Markets Day, Altman praised the o1's reasoning abilities, which he believes will create new opportunities for AI beyond what was possible with earlier GPT models. Altman described the emergence of "level 3 AI, " which refers to agentic systems that can perform tasks independently to provide optimal responses. This advancement could pave the way for "level 4" systems that can innovate. He discussed the current strengths of the GPT series, particularly the GPT-4o, which is multimodal but operates like earlier models by processing information token by token. He highlighted the transition from fast, intuitive decision-making (system 1) to more logical and deliberate reasoning (system 2) with the o1 model.

Altman expressed that o1 is at an early development stage, comparable to GPT-2, but anticipates significant advancements in the coming months, with upgrades transitioning it from o1-preview to a complete version. The o1 model represents a paradigm shift in AI, moving from simple prompt-response interactions to a more human-like problem-solving approach through a chain-of-thought process. This change aims to reduce issues like hallucinations and inaccuracies common in earlier models. Altman emphasized that fields like healthcare and education could greatly benefit from these reasoning models, envisioning personalized learning and advancements in scientific discovery. He underscored OpenAI's ongoing commitment to deep learning and their aim toward artificial general intelligence (AGI). Finally, Altman announced that the full version of o1 is expected to launch soon, followed by advanced iterations, though it's yet to be determined if subsequent models will incorporate the agentic capabilities he envisions.