A few days ago, OpenAI announced a major change in how ChatGPT works. From now on, the American startup wants the generative AI to adopt a form of "intellectual freedom", no matter "how difficult or controversial a topic may be". In short, ChatGPT should no longer hide behind a single ideology or the refusal to answer a controversial topic.
From now on, the chatbot will put forward multiple perspectives on the most difficult topics. By reducing censorship, OpenAI wants ChatGPT to be able to be more neutral. This approach goes against the way OpenAI has programmed AI safeguards so far. OpenAI has indeed always opted for a cautious approach. Between the lines, Sam Altman’s company admits to having built “arbitrary restrictions” into ChatGPT since its inception.
No Idea Is Off-Limited
In concrete terms, OpenAI will let ChatGPT assert that “all lives matter”, just as much as “black lives matter”. The AI will no longer systematically refuse to provide an answer, or no longer steer its discourse in a particular direction. The AI “can remain neutral on topics that some consider morally wrong or offensive”, which will avoid “shaping humanity”, instead of helping it.
In the process, OpenAI made a point of removing ChatGPT’s warning messages that flagged content that might violate its terms of use. According to Nick Turley, OpenAI’s product manager, many of the warnings were “unnecessary.” However, ChatGPT will still steadfastly refuse to answer questions about criminal activity. You won’t be able to ask the AI for a tutorial on how to make a bomb or produce drugs.
Is OpenAI Trying to Please Donald Trump?
When contacted by TechCrunch, an OpenAI spokesperson said that the idea of “giving users more control” is a “long-held belief of OpenAI.” The start-up assures that it has not changed its approach following the election of Donald Trump. The new president has fiercely opposed strong moderation on digital platforms, such as social networks. Several tech giants close to the Trump clan had also castigated OpenAI's approach, considered liberticidal and likely to manipulate public opinion.
To please the new tenant of the White House, Meta has considerably relaxed its moderation policy. To prioritize "freedom of expression", Mark Zuckerberg's group has simply removed fact-checking. This is the case for several other Silicon Valley giants, following in the footsteps of Elon Musk.
Source: OpenAI
0 Comments