ChatGPT has broken again, this time it looks like it is having a stroke.
Several users on Tuesday reported that OpenAI's chatbot spouting gibberish words when answering simple inquiries.
While not the first time it happened, it shows that the AI technology we have today will not replace humans anytime soon. Well, at least not in the near future.
Here are some of the probable reasons why ChatGPT seems like to be suffering from a stroke:
Error in Database Training
The most plausible reason is that there was an error in the recent database OpenAI is using to train ChatGPT.
With publishers, authors, and news outfits clamping down on how much data OpenAI can collect to train ChatGPT, errors like this can become common.
OpenAI cited similar reasons before when the chatbot was called out for being "lazy." It will be no surprise if the same problem is surfacing here.
ChatGPT 'Memory' Glitch
It is worth noting that the error was first reported following the company's announcement of improving ChatGPT's performance by giving it a "memory."
The new feature was touted as a way for the chatbot to remember things "you discuss across all chats, saving you from having to repeat information and making future conversations more helpful."
With text AIs generating output based on the information provided from their cloud servers, the addition of the "memory" could possibly corrupt its function as it collects information from multiple sources.
However, there are no clear indications yet that the recent update has really caused the chatbot to suffer a digital stroke.
Related Article : OpenAI Gives ChatGPT Memory, Assures Users Still in Control
Server Issue
A much more concrete, albeit less exciting, reason is that there was a problem with OpenAI's servers.
Issues like these usually happen in cloud computing companies providing services for millions of people at the same time.
Similarly, these problems often take hours to get fixed or the platform issues an update to resolve them.
So far, OpenAI has already been notified of the error and said that it will be "investigating reports of unexpected responses from ChatGPT." The error was determined to be a bug.
