Despite ChatGPT’s many amazing capabilities, it has a major weakness: a lack of information about current events. However, it appears that OpenAI may be quietly working on a solution.
When OpenAI first unveiled ChatGPT nearly a year ago, the AI chatbot only had knowledge of information that occurred before September 2021 since the data it was trained on only covered that time range.
Also: Best AI chatbots
However, some users recently took to X (formerly Twitter) to share that they’ve noticed an expansion in the time frame of knowledge the chatbot possesses.
In the example above, the user asked ChatGPT Plus with GPT-4 what the knowledge limit was, and he responded with “September 2023,” making ChatGPT knowledge scope very recent.
Other ChatGPT Plus users in the thread got the same response from the AI-powered chatbot regarding the scope of its knowledge. However, when asked about current events that occurred during the time period he claims to be aware of, ChatGPT doesn’t seem to have the answers.
ZDNET decided to test it and asked both ChatGPT Plus with GPT-4 and ChatGPT Standard with GPT-3.5 about the scope of their knowledge. Either way, ChatGPT responded in January 2022.
While this isn’t as recent as what other users have been getting, expanding knowledge to January 2022, if effective, would still represent significant progress for the chatbot since its initial cut in September 2021.
Also: GPT-3.5 vs GPT-4: Is ChatGPT Plus worth its subscription fee?
Similar to the experience described by X users, although ChatGPT claimed to have knowledge of information as of January 2022, it was unable to answer questions about events that occurred in December 2021.
When asked when the first Omnicron death occurred and who won the 70th Miss Universe pageant, ChatGPT said it was unable to access that information since it occurred after the January 2022 deadline — which is incorrect — and to check the website for the latest information.
OpenAI’s ChatGPT FAQ page, last updated last week, states that ChatGPT “has limited knowledge of the world and events after 2021 and may occasionally produce malicious instructions or biased content.”
Also: 8 ways to reduce ChatGPT hallucinations
Why ChatGPT claims to have knowledge it does not have is unclear. ZDNET has reached out to OpenAI for comment.
However, what it does point out is the need to verify the information you get from ChatGPT because, like any other generative AI model, it is susceptible to hallucinations.