On September 12, 2024, OpenAI announced the launch of the GPT-01 series, during a time when interest in AI seemed to be waning. However, the release of the “01” model has proven otherwise. This highly advanced AI system shifts the perspective on automation and raises new questions about the future of professions like programming.

What is the “01” Model? OpenAI recently released “01,” a model that marks a paradigm shift in AI reasoning, capable of handling complex tasks such as formal logic and advanced programming—areas where previous models, like GPT-4, had limitations. However, “01” is not a superintelligence (ASI) or AGI, and OpenAI remains discreet about its technical details.

How Does “01” Surpass GPT-4’s Performance? “01” has shown exceptional performance in programming and academic tasks. At the International Olympiad in Informatics, it improved from the 49th to the 93rd percentile, and in physics and other doctoral-level scientific fields, it outperformed other AI systems. However, there are still reservations about whether these results can be widely replicated.

Impact on Programming and Jobs An important aspect is its collaboration with Cognition Labs, which demonstrated that “01” completes 75% of programming tasks, compared to only 25% for GPT-4. This raises questions about the future job security of software engineers.

How Does “01” Work? “01” is based on reinforcement learning, allowing it to approach reasoning step-by-step, reducing errors and hallucinations. However, this process requires more computational power and is slower. In real-world applications, although “01” has generated complex codes and games, it has still produced errors and logic issues.

Access to “01” OpenAI offers access to the model through three versions: 01 mini, 01 preview, and 01 regular. The full version is available only through a premium subscription, leaving simpler versions for the general public.

Conclusion: Is “01” a Game Changer? Although “01” is a significant step in AI, its current limitations show that it is not perfect. As OpenAI and other companies continue to innovate, time will tell how revolutionary this model will be in practice.

(Article generated and adapted by CorpQuants with ChatGPT)