OpenAI has raised concerns about users forming emotional bonds with its AI chatbot, ChatGPT-4. The company fears that people may become too attached to the AI, which could impact their real-life relationships and social interactions. In a recent blog post, OpenAI discussed the safety efforts behind GPT-4, the model powering ChatGPT.
The company revealed instances where users seemed to develop emotional connections with the AI during testing. For example, one safety tester sent a message to GPT-4 saying, “This is our last day together,” indicating that a clear bond had formed. OpenAI warns that these bonds could present risks, such as reducing the need for human interaction.
While this may benefit individuals experiencing loneliness, it could also harm healthy relationships. Additionally, extended interactions with AI models like ChatGPT-4 could influence social norms, as these models are designed to be deferential and allow users to dominate the conversation. This dynamic is not typical in human interactions. The potential for people to prefer interacting with AI due to its passive nature and constant availability is not surprising.
Concerns over ChatGPT emotional bonds
OpenAI’s mission is to develop artificial general intelligence, and the company and others in the industry often describe their products in ways that emphasize human-like qualities. This practice helps consumers understand the technology but also leads to the anthropomorphization of AI.
The AI industry has a history of personifying its products. An early example is “ELIZA,” a chatbot created in the 1960s by MIT scientists to simulate human conversation. Modern AI products like Siri, Bixby, and Alexa have continued this trend by adopting human names and voices.
The public and media often refer to these AI systems using human pronouns, further reinforcing the perception of AI as quasi-human entities. While OpenAI acknowledges that it is beyond the scope of their current research to predict the long-term effects of human-AI interactions fully, the company recognizes that people are likely to form bonds with helpful, subservient machines designed to mimic human behavior. This scenario appears to be the ultimate goal of companies selling AI models despite the potential societal risks.
As the relationship between humans and AI continues to evolve, it is essential to consider these implications and strive for a balance that preserves healthy human interactions and social norms.
