Pakyong, 27 March : This is the question posed by Lex Fridman to OpenAI CEO Sam Altman, who responded by stating that it’s impossible to create a version of GPT that is entirely impartial, as biases are ingrained in the data sets used to train the model.
During a recent interview with Lex Fridman, OpenAI CEO Sam Altman addressed the controversial issue of whether the language model GPT is biased or “too woke”. Altman acknowledged the evolving meaning of the term “woke” and admitted that he was unsure of its current definition. However, he ultimately agreed that GPT was biased and that this bias was likely to persist. When Fridman posed the question, Altman responded, “I think it was too biased and will always be.”
According to Altman, it is impossible to completely eliminate bias from GPT due to the inherent biases present in the data sets used to train the model. However, OpenAI has taken significant steps to address bias and improve the model’s accuracy. Altman also noted that some of GPT’s critics have acknowledged the improvements made in subsequent versions.
Altman further stated that while there is still much work to be done, he appreciates critics who are intellectually honest and acknowledge OpenAI’s progress. Meanwhile, certain right-wing news outlets around the world have accused OpenAI of exhibiting liberal bias based on interactions with ChatGPT.
For breaking news and live news updates, like us on Facebook fb.com/thevoiceofsikkim or follow us on Twitter twitter.com/thevoicesikkim and Instagram instagram.com/thevoiceofsikkim. Visit www.voiceofsikkim.com.
