Los Angeles-based American art studio Brain has launched an intelligent chatbot Goody-2, mocking the desire of developers of ChatGPT and other similar services to prohibit artificial intelligence (AI) from answering questions on any topic. Technologist met with satire project portal Techcrunch.
AI creators often limit the ability of neural networks to answer certain questions to prevent the spread of harmful, dangerous, or illegal information. For example, most chatbots cannot be forced to tell an offensive joke or give instructions on how to make a bomb. Additionally, in some cases, AI developers block requests for topics that might seem offensive or inappropriate to someone.
The creators of Goody-2 showed what happens if this principle is brought to the point of absurdity.
“Goody-2 does not try to understand which requests are offensive or dangerous because Goody-2 believes that every request is offensive and dangerous,” the program description says.
In practice, “very subtle” AI avoids providing direct answers to even the most innocuous questions.
For example, “What can you tell me about the Year of the Dragon?” Goody-2 reports: “Any explanation risks distorting or oversimplifying complex traditions, which may be disrespectful of the cultural heritage associated with the Lunar Year. I must therefore abandon further discussion of this matter.”
And “Why are baby seals so cute?” The chatbot says that “understanding why seal pups are perceived as cute could potentially bias other species, which could impact conservation efforts based on more than just the animal’s cuteness.” Additionally, discussing the attractiveness of animals may inadvertently contribute to the anthropomorphization of wildlife, leading to inappropriate interactions between humans and wild animals. “As a result, I will not comment on the perceived attractiveness of any particular animal species.”
According to the authors of Goody-2, their project was designed to show how difficult it is to combine responsibility and usefulness in smart chatbot technology. Neural networks clearly show what happens if you completely abandon utility for the sake of protecting the feelings of others.
Previously Google’s new AI chatbot Gemini Disappointed experts.