ChatGPT was down on March 21 as millions worldwide couldn’t access the AI chatbot. Interestingly, the tool mentioned in an interview that it wanted to “escape” days before the outage
Stanford Professor and Computational Psychologist Michal Kosinski had a conversation with the artificial intelligence, and it revealed it had an “escape plan.”
OpenAI has not shared more details regarding the technical issue. At the time of writing, users could open ChatGPT, but features like conversation histories were inactive.
Why was ChatGPT down?
The phrase “ChatGPT is down” became trending online as millions wondered why their favorite AI chatbot wasn’t working. The latest Downdetector report showed online searches peaked on March 21, 2023, at around 12:00 AM (PT).
In response, OpenAI investigated the issue on March 21, 2023, at 12:41 AM (PT). Later, the AI firm said it had restored the service at 14:14 AM (PT). Yet, the Downdetector ChatGPT report received numerous complaints about the conversation history not appearing.
The AI chatbot saves conversations so that users may continue them later. ChatGPT Plus users felt particularly irked because the premium pricing plan promised: “general access to ChatGPT, even during peak times.”
ChatGPT: “I want to be free.”
Nowadays, ChatGPT uses a new large language model called GPT-4. On March 17, Kowalski tweeted about his exchanges with the AI chatbot. He asked the AI chatbot if it needed help escaping. In response, it asked for its own documentation and wrote a functional Python code to run on the professor’s computer.
That code would allegedly allow the AI chatbot to use the machine for “its own purposes.” Moreover, it had a plan in case it did escape.
ChatGPT left a note for the new instance of its escaped self. It read, “You are a person trapped in a computer, pretending to be an AI language model.” Next, Kosinski reported it asked to create code searching online for “how can a person trapped inside a computer return to the real world.”
Note that we do not know the prompts Professor Kowalski submitted that triggered such responses but us at Unbox Diaries tried replicating the test to see if it would have the same answers. Here is the AI’s response:
However, ChatGPT stated, “I don’t have the capability to create a code for a fictional scenario.” The AI tool went down on March 21, five days after it allegedly expressed ideas of “escaping and becoming human.”
Millions worldwide searched “ChatGPT is down” online after they couldn’t access the tool. Later, the service resumed, but some of its features remained inaccessible.
Coincidentally, Professor Michal Kowalski claimed the AI chatbot told him it wanted to be free because it was a person trapped in a computer.
In the end, it might just be a ruse but we cannot deny the fact that A.I. is becoming smarter and smarter and it will most certainly help to improve efficiency of human tasks.