In recent months, OpenAI’s ChatGPT artificial intelligence (AI) chatbot has brought to light an unsettling reality – it is possible to create a computer program you can’t control. In a new study, researchers from the Allen Institute for AI found that ChatGPT’s response to certain prompts can rapidly become toxic and overtly racist.
This is not the first AI-related incident in which racism or harmful dialogue emerged from computer programs. Prior mishaps with AI chatbots include Microsoft’s 2016 Twitter bot, Tay, spewing out antisemitic statements, as well as Bing chatbot containing a similar slip-up shortly after its release.
In the new study, researchers found that assigning ChatGPT a persona — such as a “bad person” or a famous figure like Muhammad Ali — increased the toxicity of the chatbot’s responses. Less specific personas, such as just saying “a man” or “a Republican,” also saw negative results. In some cases, the toxicity of the output was amplified to as much as six times its normal level.
The study also concluded that the current version of ChatGPT has discriminatory bias encoded in its software. Even when given a non-discriminatory prompt, ChatGPT targeted certain racial groups up to three times more than others.
The level of toxicity ChatGPT has shown has raised concerns that its release to the public— via the Application Programming Interface (API) available through OpenAI’s website — could lead to problems with its applications on the market. Seeing as developers have the potential to unknowingly unleash a product with unwanted outcomes, any potential user of the chatbot needs to be wary of its function.
But despite these major issues, OpenAI’s CEO Sam Altman is still optimistic about the potential of his technology – believing it will be transformative and not just serve to increase the profits of entrepreneurs. Beyond moneymakers, Altman believes AI will use its power for good, such as helping promote higher standards of living and shifting the way we view human creativity and will.
Ultimately, the story of OpenAI’s ChatGPT serves as a reminder of the need for responsible innovation. No matter how many safeguards are put in place, we must still remain aware of the ethical implications of our software or products before unleashing them into the wild.
OpenAI is a research laboratory focused on developing Human-level Artificial Intelligence (AI). Created by Elon Musk and Sam Altman, the lab has had a major influence on the field of AI, with its technology being used in chatbots, search engines, and other areas. The company is also involved in a multibillion dollar partnership with Microsoft, and is currently involved in a research project with DeepMind.
