Recently, the renowned programming competition platform Codeforces made a significant decision – to ban the use of artificial intelligence (AI) in programming contests. As AI programming capabilities rapidly improve, Codeforces founder Mikhail Mirzayanov stated that this decision was necessary because neural networks have reached “new heights that cannot be ignored.”
This new rule primarily targets contest participants, explicitly prohibiting the use of AI systems like GPT, Gemini, Llama, and Claude during competitions. Mirzayanov mentioned that AI’s progress in programming is not to be underestimated, especially with OpenAI’s recently released o1 model performing impressively in programming contests. In simulated Codeforces competitions, o1 achieved an Elo rating of 1807, surpassing 93% of human participants and even approaching the “master” level in a live competition.
However, Codeforces is not implementing a complete ban on AI usage. The new rules allow participants to use AI within certain limits, such as for translation tasks or minor grammatical and coding suggestions. But using AI to generate core logic or algorithms is strictly forbidden. Nevertheless, enforcing this ban may prove challenging, as Codeforces primarily relies on the integrity of its participants.
In this context, some users have expressed dissatisfaction with the decision, arguing that current technology allows participants to easily modify AI-generated code to make it appear unique without truly understanding the code’s essence. They point out that future programming competition platforms will actually depend on the trust of participants, and efforts to combat AI models may be destined to fail.
Similarly, AI has raised comparable issues in other competitive intellectual games like chess and Go, where AI capabilities have surpassed human abilities. However, in these competitions, human players directly compete against each other, preserving the spirit of competition. This is more difficult to ensure in anonymous online programming contests.
To maintain fair competition, Codeforces stated that they will closely monitor AI technology developments and adjust rules as needed to balance fair competition and AI-assisted learning. Meanwhile, renowned programmer George Hotz highly praised the o1 model, considering it the first AI truly capable of programming and expressing confidence in its future.