KPMG has pulled a report from its website that claimed to analyze the effectiveness of artificial intelligence systems. The report, titled "Assessing the Reliability of Artificial Intelligence," was sent to clients and was available online for several weeks before it was taken down. The report did not provide any concrete data or conclusions on AI's reliability.
However, employees at KPMG claim that the report included "hallucinations" - or false positives - in its conclusions about the performance of AI systems. According to sources within the firm, certain sections of the report were heavily criticized by colleagues for being inaccurate and misleading. One employee told our publication that they saw a section where the authors claimed that an AI system was 90% accurate when, in reality, it was only around 10%. The employee expressed frustration at how difficult it was to get their concerns heard within the company.
KPMG's decision to pull the report has sparked controversy and raised questions about the reliability of such analyses. Many experts have long argued that AI systems are prone to errors and biases, making them unreliable sources of information. This experience is not an isolated incident - similar instances have been reported in various fields where AI research and development are being conducted.