The smarter artificial intelligence is, the more likely it is to hide its ignorance?

An artificial intelligence (AI) research team at the Polytechnic University of Valencia, Spain, has found that as large language models become larger and more sophisticated, they tend to be less likely to admit to users that they don't know the answer.

The smarter the AI, the less likely it is to admit to users that it doesn't know the answer. (AI illustration)

In the study, published in the journal Nature , the team tested the latest versions of three of the most popular AI chatbots on responsiveness, accuracy, and the ability of users to spot incorrect answers.

To test the accuracy of the three most popular LLMs, BLOOM, LLaMA, and GPT, the team asked thousands of questions and compared the responses received to previous versions of the same questions. They also varied the topics, including math, science, word problems, and geography, as well as the ability to generate text or perform actions like sorting lists.

The study revealed some notable trends. The overall accuracy of chatbots improved with each new version, but still declined when faced with more difficult questions. Surprisingly, as LLMs became larger and more sophisticated, they tended to be less open about their ability to answer correctly.

In earlier versions, most LLMs would explicitly tell users when they couldn’t find an answer or needed more information. In contrast, newer versions tend to guess more, resulting in more answers overall, both correct and incorrect. More worryingly, the study found that all LLMs still occasionally gave incorrect answers even to easy questions, suggesting that their reliability remains an area of ​​improvement.

These findings highlight a paradox in the evolution of AI: while models are becoming more powerful, they may also become less transparent about their limitations.

This poses new challenges in the use and trust of AI systems, requiring users to be more cautious and developers to focus on improving not only the accuracy but also the “self-awareness” of models.

Source: https://baoquocte.vn/cang-thong-minh-tri-tue-nhan-tao-cang-co-xu-huong-giau-dot-287987.html