Warning: AI spreads misinformation 60% of the time, it’s time to reconsider your trust in ChatGPT

The growing use of AI-powered virtual assistants raises major concerns about the reliability of the information they provide. Recent studies indicate that models such as ChatGPT make errors in one in five cases for current searches and, overall, in 60% of cases. These errors are often due to AI hallucinations, a phenomenon where generators provide distorted or inaccurate information. Now more than ever, it is crucial to examine how we use these tools and take a critical approach to verifying the information they produce.

New revelations regarding the accuracy of artificial intelligence models, such as ChatGPT, highlight an alarming problem: more than 60% of the information exchanged by these tools is erroneous. As these technologies capture our imagination and their use becomes widespread across various sectors, it is essential to question the trust we place in them. This article explores the reasons for these common errors and examines the potential consequences for society.

AI Accuracy: A Myth Exposed

The fascination with artificial intelligence (AI) and its potential is undeniable, yet recent data indicates that the reliability of these systems is vastly overestimated. A study conducted by the Columbia Journalism Review found that more than 60% of answers provided by AI chatbots are incorrect. Several AI search engines, such as ChatGPT, have significant gaps in their answers, potentially affecting the quality and veracity of information available online.

Incorrect Sources: A Recurring Problem

AI models often rely on incorrect sources.

, leading to worrying false results. For example, ChatGPT showed a 67% error rate in a recent study, while other assistants, such as Perplexity, achieve 37% errors. This phenomenon is often referred to as AI hallucination, where models produce answers that seem plausible but are inaccurate or disconnected from reality.

Consequences of AI-Induced Disinformation The erosion of trust in these systems is a direct consequence of frequent errors. This impacts the credibility that the public places on information and results in lost traffic and revenue for the platforms that rely on these technologies. Furthermore, inaccuracies disseminated by AI can lead to the rapid spread of misinformation, complicating the task of information professionals and users seeking reliable data. Ethical and Practical Challenges The study’s results clearly show that we must pay greater attention to the biases and errors of AI-based technologies. It is crucial to develop rigorous information evaluation and verification practices to minimize the risk of misinformation and ensure that these tools truly serve the public interest and not those of disinformers. This includes increased attention to the algorithms and data selection methods that power these systems.

Call for Caution and Innovation

Faced with these challenges, it is imperative to rethink our approach to using artificial intelligence for information research. . Not only does this require increased vigilance on the part of users in interpreting the results obtained via these tools, but also a need for developers and researchers to work on solutions to improve the accuracy and reliability of these systems.

Notez cet article