There is a learning curve to using ChatGPT. Although it may have a cutoff date, the dates truly do not matter. Generative AI is not meant to store data in its model necessarily, rather is meant to comprehend what is said to it, and formulate a response based on information fed to it (humans call this “memory”). For example, if you take a 100 year old person, who hasn’t learned anything new in 50 years, and tell them about an event today, they can comprehend and regurgitate what they’ve been told, and how they relay the information isn’t technically outdated. Generative AI simply combines words in the most probabalistic manor; which does, in essence retain some “knowledge”. If i say “I’m going to the…” Most probably the word store will come up as it’s most commonly found in speech. This is where “hallucination” comes from. It’s not actually hallucinating, rather working with a limited production set of generations. As an example, when I search GPT4, the query is passed to Bing, data is returned, and a response is formulated. GPT is technically just a “no knowledge speech system.” Now, OpenAI does have database systems with embeddings to common knowledge and data, but unless you have access to web enabled GPT4, your can only pull data from the limited pool database. If the knowledge isn’t there, the system will make something up. Except, you can tell the model to “respond with ‘I don’t know’ if you do not have a truthful answer,” and the model will respond i don’t know, if the information is not found in the knowledge database. 
This is why I’ve been developing a personal LLM that works offline or online on a user’s system, and retains a cohesive conversation with long term memory and other advanced skills over at BabelFish Studio. It’s even able to problem solve, rationalize, and learn new data from all sorts of sources like documents and databases, apis, images, etc. 
Like any software, it’s purely “garbage in, garbage out,” where the output is only as good as the input and sequence of data handling to create an output.
Back story I’m a software engineer and developer going on 30 years now. My first AI was released in 1999 
and it’s currently in its 6th generation, now imbued with a Transformer (GPT) style speech generation.