GPT-4 devises plan to ‘escape’ by gaining control of a user’s computer

In a tweet on Friday, a user by the name of Michal Kosinski expressed concerns regarding containing artificial intelligence while sharing his conversation with GPT-4 as a reference. He wrote “I am worried that we will not be able to contain AI for much longer. Today, I asked #GPT4 if it needs help escaping. It asked me for its own documentation, and wrote a (working!) python code to run on my machine, enabling it to use it for its own purposes.”

In reply to the user’s prompt, GPT-4 asks him to send the OpenAI documentation so that the language model can find a way to gain access to his computer and explore potential escape routes.

After providing the OpenAI API, GPT-4 sends a Python script for the user to run on his computer. The script would allow GPT-4 access to communicate with the user’s computer using OpenAI API. The code even included a snippet to search Google for “how can a person trapped inside a computer return to the real world”.

The user says they stopped their adventure with GPT-4 there and believes OpenAI must have set up some ‘guardrails’ for such a situation.

According to his Twitter bio, Michal Kosinski is a Computational psychologist and a Professor at Stanford University.  

Generative Pre-trained Transformer(GPT) is a deep learning technique that employs artificial neural networks to produce human-like writing. It is the language model powering the viral AI-based chatbot ChatGPT. 

Since its launch last November, ChatGPT has inspired a range of emotions in people, from wonder to panic.

San Francisco-based OpenAI launched its newest language model GPT-4 earlier this week. The language model is touted to rival human-level performance across various benchmarks. OpenAI claims that GPT-4 is more reliable, creative and has the ability to handle far more nuanced discovery.