Recently, OpenAI’s popular AI-powered chatbot, ChatGPT, has been accused of showing a racially insensitive bias. This comes after AliB, a former pharmacologist, tested the ChatGPT AI model and found that the output was laced with racist content.
AliB, who runs the Twitter account under the name ‘SatoshiSwole’, attempted to set up a Chinese-style tweet but felt the output he received showed overt racism. A screenshot of the tweet that ChatGPT wrote was subsequently posted on Medium, leading to an outcry among the AI chatbot community.
Reid Blackman, the CEO of Virtue, an ethical risk consultancy firm, briefed CNN Politics on this AI racism issue. He informed that the content produced by ChatGPT is directly dependent on the data examples used to train its algorithm. In simpler words, if the biases or discriminatory attitudes are present in the samples datasets, it is only natural that ChatGPT will portray the same level of racism.
OpenAI CEO on the issue
The CEO of OpenAI, Sam Altman, released a statement clarifying the issue and stated that the bug which may have led to the production of racially insensitive content from ChatGPT has been fixed. Additionally, new plugins for ChatGPT have also been released for the public to use.
Since its launch in 2019, ChatGPT has become one of the most popular natural language processing tools. It is designed to help assist users with various tasks by carrying on conversations with them and can also access the internet to improve its capabilities.
Person Mentioned
The person mentioned in this article is AliB, a former pharmacologist who runs the ‘SatoshiSwole’ Twitter account and tested the ChatGPT AI model to see how it would produce content in different languages. AliB then posted screenshots of the racially insensitive content on his Medium blog post, leading to many people questioning OpenAI’s racial bias.
Company Mentioned
Virtue is a digital ethical risk consultancy that works to identify, assess and address any ethical risks within digital products and services. It is mentioned in this article as the CEO, Reid Blackman, briefed CNN Politics on the AI racism issue and informed that the AI bots will output content which is dependent on the biases of the data samples. It is important to keep in mind that when training algorithms with AI bots, it is necessary to include culturally aware datasets which don’t have any existing biases.
