AI chatbot, can we trust it? Overcoming reliability is the biggest challenge.

Author: DigitalToday | Source: DigitalToday | Read the full article

In recent discussions about artificial intelligence, a significant question has emerged: Can we trust AI chatbots? A study highlighted by IT media outlet 9to5Mac examined the reliability of popular AI chatbots like ChatGPT and Gemini. The research raised concerns about the accuracy of these chatbots when used as alternatives for web searches, revealing that they often provide incorrect information with unwarranted confidence.

The study involved testing eight AI chatbots to see how well they could find accurate quotes and details from articles. Unfortunately, the results were disappointing, with most chatbots delivering either partially or completely incorrect answers. On average, these AI systems were accurate less than 40% of the time, with Perplexity being the most reliable and Grok-3 the least.

One of the key takeaways from the research is that chatbots tend to give inaccurate answers to questions they cannot answer correctly. Interestingly, paid chatbots were found to be more confident in their incorrect responses compared to free ones. This study suggests that while AI chatbots can be useful for inspiration and generating ideas, they should not be relied upon for accurate information.

[Read More]