The Irony: Hallucinations in Papers at NeurIPS, a Leading AI Conference

In the era of generative artificial intelligence, even the most prestigious scientific conferences are not immune to problems. A study conducted by the GPTZero startup revealed instances of "hallucinations" — AI-generated fictitious citations — in scientific papers presented at NeurIPS, one of the leading conferences in the field of artificial intelligence. The irony of the situation is obvious: a conference dedicated to cutting-edge achievements in AI has itself become a victim of its side effects.

GPTZero, a company specializing in detecting AI-generated text, conducted an analysis of a corpus of papers accepted at NeurIPS. The results were alarming: a number of papers contained references to non-existent sources, apparently created by large language models (LLM). This calls into question not only the quality of individual papers, but the entire system of scientific peer review.

The problem of hallucinations in LLMs is well known. Models trained on vast amounts of data sometimes "fabricate" facts or citations to fill gaps in their knowledge or make text more persuasive. In the context of scientific papers, this can have serious consequences, undermining trust in research and making it difficult to verify results.

The discovery of AI-generated citations in NeurIPS papers raises several important questions. First, it demonstrates the need for more careful examination of scientific papers for the use of artificial intelligence. Reviewers must be more attentive to details and double-check cited sources. Second, it underscores the importance of developing tools to detect AI-generated text and implementing them in the scientific publication process.

Moreover, this incident prompts reflection on the future of scientific research in the age of AI. On one hand, artificial intelligence can become a powerful tool for scientists, helping them analyze data, generate hypotheses, and write papers. On the other hand, it is necessary to consider the risks associated with the use of AI and take measures to prevent abuse.

Ultimately, the discovery of hallucinations in NeurIPS papers is a wake-up call. It reminds us that artificial intelligence is merely a tool, and its use requires responsibility and critical thinking. The scientific community must develop clear rules and procedures for working with AI to ensure the quality and reliability of scientific research.