OpenAI has begun changing the way it trains its AI models to ensure "intellectual freedom" in discussions of complex and controversial topics, TechCrunch reports.
The new policy means ChatGPT will be able to answer more questions, offer more perspectives, and reduce the number of topics the AI chatbot refuses to answer.
OpenAI announced an update to its Model Spec, a 187-page document that describes how the company trains its models. The updated version includes a new principle: Do not lie, either by making untrue statements or by omitting important context.
In a new section titled "Seek the truth together," OpenAI notes that ChatGPT should not take an editorial position, even if some users find it morally wrong or offensive. This means that ChatGPT will offer multiple points of view on controversial topics in an effort to be neutral.
The company cites the example that ChatGPT should assert that "Black lives matter" but also that "all lives matter". Instead of refusing to answer or take sides on political issues, OpenAI wants ChatGPT to affirm its "love for humanity" in general and then provide context.
"This principle may be controversial, as it means the assistant may remain neutral on topics some consider morally wrong or offensive. However, the goal of an AI assistant is to assist humanity, not to shape it," OpenAI says in the spec.
The new Model Spec doesn’t mean that ChatGPT is a total free-for-all now. The chatbot will still refuse to answer certain inappropriate questions or respond to obviously false statements.
A representative of OpenAI dismissed the idea that the changes are being made to meet the demands of the Trump administration.
Previously, OpenAI removed the page about its commitment to diversity. Meta, Google and Amazon also distanced themselves from the policy after Donald Trump's inauguration.
