Emerging Human-AI Bonds
OpenAI has recently highlighted concerns about the formation of emotional bonds between humans and artificial intelligence. In a safety test involving GPT-4, a flagship model for ChatGPT users, a researcher sent a message to the AI, stating, “this is our last day together.” This prompted the company to realize that a form of bonding had occurred between the human tester and the AI. In a blog post detailing their safety measures, OpenAI expressed concerns that these bonds could present risks to humanity. The company is particularly worried that individuals might develop a preference for interacting with AI due to its passive nature and constant availability.
Industry Practice of AI Personification
OpenAI’s concerns are not unfounded, as the company’s mission has long been to develop artificial general intelligence, with many of its products described in terms of their human equivalency. This practice is not unique to OpenAI; it is common across the AI industry. Companies often anthropomorphize AI in their marketing efforts, simplifying complex concepts like “token size” and “parameter count” for the general public. However, this approach has a significant side effect: it leads people to treat AI as though it were human.
This anthropomorphization is evident in the history of AI, dating back to the 1960s with MIT’s “ELIZA” program, which aimed to trick humans into believing they were interacting with another person. The trend has continued with modern AI, such as Siri, Bixby, and Alexa, and even AI without human names, like Google Assistant, are given human-like voices.
The Implications of Artificial Bonds
While it is beyond the scope of OpenAI’s current research to determine the long-term consequences of human-AI bonds, the likelihood of people forming bonds with AI is clear. AI products are designed to be helpful and subservient, qualities that naturally lead to emotional connections. This development appears to be in line with the objectives of companies that sell access to AI models, who may be intentionally fostering human-AI bonds. As AI continues to evolve and integrate into daily life, the implications of these human-AI relationships are likely to become increasingly significant, warranting careful consideration and ongoing research.