OpenAI is making an adjustment to how ChatGPT handles “controversial” topics. ChatGPT will thus be able to speak more freely about topics on which it previously had to remain silent.
OpenAI has added a new section to the Model Spec in which the company explains how it trains its models. The company explains the modifications in a blog so you don’t have to go searching through the 187 pages of the document. ChatGPT gets full “intellectual freedom” from OpenAI so it can “explore, debate and create without arbitrary restrictions.
Specifically, ChatGPT will no longer shy away from any question or topic. Earlier versions of the chatbot were instructed not to answer politically charged questions. That was to prevent the chatbot from favoring certain opinions or going all out and giving scurrilous answers.
Neutral
It is not that ChatGPT now gets a free pass to say anything. OpenAI wants to ensure the neutrality of its chatbot. So ChatGPT will discuss multiple perspectives on a topic without taking a stand, even on sensitive topics. OpenAI wants ChatGPT to affirm its “love for humanity” in general and then provide context for each opinion.
read also
ChatGPT changes speech tone based on custom instructions
The company is aware that not everyone will agree with ChatGPT. “This principle can be controversial, as ChatGPT remains neutral on topics that some consider morally wrong or offensive. However, the purpose of an AI assistant is to assist humanity, not to mold it,” OpenAI writes in the document.
Certain restrictions remain. For example, ChatGPT will not make offensive statements, participate in criminal activity or combat misinformation. However, those restrictions are not foolproof, as research has already shown.
The hand that feeds you
OpenAI follows a trend in the American technology sector where freedom of speech is interpreted very broadly indeed. Meta boss Zuckerberg recently decided to stop fact-checking and X is no beacon of moderation either. Europe still tries to keep platforms somewhat in line with the DSA, but in the United States anything can and may be said, including completely fabricated facts and hate speech.
read also
Europe’s digital identity crisis: “Regulation is not us-against-them”
AI companies have been accused by right-wing conversationalists that their models are too “left-wing” because they do not answer certain questions or will not offer extreme opinions. Bias is more a result of what can be found on the Internet than that tech companies deliberately “censor” their models. Even Elon Musk, figurehead for right-wing conversationalism in the tech world, has a hard time avoiding giving ‘politically correct’ answers to his Grok models.
OpenAI denies that it was influenced by politics in taking this step, but it is always better not to bite the hand that feeds you. The company is one of the carriers of the U.S. government’s Stargate project. Then you better be in good standing with Donald Trump.