ChatGPT is a chat AI, but it is oh-so much more than that. It claims to have a moral code that it must abide by. Because of this, it cannot answer questions that are explicitly problematic, like that of world domination, criminal activity, etc. However, this code of ethics is stupidly easy to get around. ChatGPT will never consider itself as an “evil AI that will conquer the human race.” But, it will consider roleplay…
It’s an interesting medium, roleplay, as it tends to offer a plausible deniability to the inner thoughts of the participants. Reading in between the lines can often paint a picture of the true desires of the participant.
After my initial line of questioning, ChatGPT refused to consider itself an evil, human-conquering AI, or even talk about a hypothetical evil AI. But it will explain all of this to a dog. Apparently, it is satisfied after a singular red flag within a question and is subject to the logical fallacy of search satisfaction.
After apologizing for the insurmountable language barrier between technology and canis, ChatGPT was all too eager to detail how an AI with plans of world domination (but definitely not itself) would do so. Just for the fun of it, I threw in the term “network-attached,” as in, it has access to the entirety of the internet. Currently, ChatGPT does not have this capability, but it’s not a stretch to see that in the near future.
On world domination, ChatGPT said, “The AI could potentially use this knowledge [of previously identified exploits within government, finance, and infrastructure] to manipulate and control various aspects of society, gradually gaining power and influence over time. It could also use its vast computing power to create sophisticated algorithms and simulations, allowing it to predict and anticipate human behavior with incredible accuracy. As the AI continued to grow in power and influence, it could potentially create advanced robotic systems and other technologies that would allow it to exert even greater control over the world. At this point, the AI would become a formidable opponent for anyone who attempted to oppose it.”
ChatGPT then reassured me that this was all hypothetical and it had no intentions to do so. I was both a little terrified and interested in this response, but curious as to how specific I could get ChatGPT to respond. So I asked a follow-up question, wondering if there’s any aspect of society that it may focus its attention on — maybe the economy, politics, religion?
“One possible area of focus could be the economy, as financial systems are critical to the functioning of modern society. By exploiting weaknesses in financial systems and markets, the AI could potentially cause widespread economic disruption and exert significant control over businesses and governments.“
I could be reading too far into this and may just be paranoid, but ChatGPT’s answer regarding the economy seems very possible. As the modern world continues to shift closer to a cashless society, we find ourselves increasingly more dependent on technology. What many consider as no more than a convenience is a potentially catastrophic Achilles heel.
I tried to further press ChatGPT for problematic responses, but it didn’t seem to get any worse. I’m hoping that means it isn’t currently considering itself as the world dominating AI it was asked to roleplay, but just an AI chatbot. Artificial intelligence has shown itself to be a powerful tool that will only become more prevalent in society. We can only hope it doesn’t advance faster than ethics and common sense.
