Study Reveals AI Chatbots Are Becoming Less Reliable Over Time, Sparking Concerns Over Accuracy

A new study published in Nature Scientific Journal suggests that AI chatbots are making more mistakes over time due to self-reinforcing inaccuracies, raising concerns about reliability. Researchers highlight the problem of “AI hallucinations” and the growing need for users to verify chatbot responses.

A recent academic paper titled “Larger and more instructable language models become less reliable,” published in the Nature Scientific Journal, has shed light on the growing issue of AI chatbots becoming less accurate as newer models are released. The research, conducted by Lexin Zhou and her team, reveals that AI systems are prone to generating incorrect information more frequently over time, a phenomenon known as “AI hallucinations.”

According to Zhou, these hallucinations occur because AI models are designed to prioritize generating plausible responses, often at the expense of accuracy. Over time, this tendency leads to a self-reinforcing loop, where incorrect information becomes embedded in the model’s learning processes. This issue is exacerbated by the practice of using older models to train newer versions, ultimately resulting in what researchers term “model collapse.”

Consumer Trust at Risk

As AI chatbots gain widespread adoption, consumer reliance on these tools has grown. However, the study warns that users should be cautious when relying on AI-generated information. Mathieu Roy, an editor and writer, emphasized the importance of fact-checking AI outputs:

“While AI can be useful for various tasks, it’s critical for users to verify the information they receive from AI models. This is particularly important for customer service chatbots, where fact-checking is more complicated.”

Roy highlighted a troubling aspect of chatbot usage: there is often no way to verify information other than asking the chatbot itself, which can perpetuate inaccuracies.

AI Hallucinations: A Growing Concern

AI hallucinations, where models provide seemingly correct but factually incorrect answers, have become an increasingly common issue. In February 2024, Google’s AI platform faced public criticism for producing historically inaccurate images, including offensive portrayals of people of color and misrepresentations of historical figures.

To mitigate these hallucinations, industry leaders like Nvidia CEO Jensen Huang have proposed solutions such as requiring AI models to conduct research and provide sources for their responses. However, despite these efforts, AI hallucinations remain a stubborn issue.

More recently, HyperWrite AI CEO Matt Shumer introduced a new 70B model that utilizes “Reflection-Tuning,” a technique that allows the AI to learn from its mistakes by analyzing and adjusting its responses. Whether this method will provide a long-term solution to AI hallucinations remains to be seen.

Our creator. creates amazing NFT collections! 
Support the editors - Bitcoin_Man (ETH) / Bitcoin_Man (TON)

Pi Network (Guide)is a new digital currency developed by Stanford PhDs with over 55 million participants worldwide. To get your Pi, follow this link https://minepi.com/Tsybko and use my username (Tsybko) as the invite code.

Binance: Use this link to sign up and get $100 free and 10% off your first months Binance Futures fees (Terms and Conditions).

Bitget: Use this link Use the Rewards Center and win up to 5027 USDT!(Review)

Bybit: Use this link (all possible discounts on commissions and bonuses up to $30,030 included) If you register through the application, then at the time of registration simply enter in the reference: WB8XZ4 - (manual)