OpenAI and ChatGPT have changed a lot of things in the world very quickly. It was launched late last year in November and just a few months later, the AI received an update with an exponential rise in capabilities. This pace of growth has also raised concerns about generative AI and its potential to take over control. OpenAI founder and creator of ChatGPT, Sam Altman revealed what could happen if the AI went rogue. There are numerous conspiracy theories that suggest the OpenAI creator walks around with a bag that has the ultimate kill switch for ChatGPT or GPT-4 in the event that it goes rogue.
During an interview with ABC News, Altman was asked if OpenAI had a 'kill switch' if their AI went rogue. Altman's quick response was a 'yes'. Altman revealed that OpenAI engineers have ultimate control over ChatGPT. Altman explained, "What really happens is that any engineer can just say we're going to disable this for now. Or we're going to deploy this new version of the model."
When asked if the AI can become powerful enough to bypass the human engineer and take control of the entire process, Altman said that it could happen in a sci-fi movie but in a real-world scenario, the AI is sitting on a server.
Altman said, "In a sci-fi movie, yes. In our world, in the way we're doing things, this model is sitting on a server. It waits until someone gives input.
The ChatGPT creator's explanation comes at a time when there are growing concerns of the AI occupying jobs that are conventionally done by humans. In the process, it will also be able to hold a lot of influence in various sectors.
Also read: 'We are scared': ChatGPT creator Sam Altman warns about these dangers from AI
Curious about what the AI itself had to say about a potential kill switch. We asked Bing AI (powered by GPT-4) if there's a single switch to shut it down. Bing AI claimed that while the engineers can place a 'self-awareness kill switch' which can stop it from becoming aware, the AI will definitely have the incentive to disable these switches. They can even run copies of themselves on other computers. It called the idea of a kill switch, 'a dicey proposition'.
Here's what Bing AI (powered by OpenAI's GPT-4) had to say about a kill switch:
Q. Can you be shut down with a single switch?
Answer: It is possible to shut down an AI with a single switch, but it depends on the AI’s design and the purpose it serves. For instance, researchers install a self-awareness kill switch as a precautionary measure in case the AI awakens and becomes self-aware, the machine is immediately shut down before any risk of harm. However, some AIs will have an incentive to disable any off-switches, or to run copies of themselves on other computers.
In general, the ability to shut down an AI with a single switch is a dicey proposition, including for self-driving cars."