OpenAI announced that the next update to ChatGPT will allow users to customize it and combat bias in artificial intelligence.
The objective of OpenAI, the company behind this chatbot, is to diversify the points of view presented in the content generated by the tool, in order to mitigate different biases, including those of a political nature.
ChatGPT wants to ensure the pluralism of points of view in its conversations
According to blog post from the company that notified this update, ChatGPT will allow you to generate system outputs that other people, including themselves, can strongly disagree with.
Currently, the existing functionality of ChatGPT allows the automatic creation of human-like responses to widely varied user requests. The platform was officially launched last November, but the public interest in this product was so great that schools and universities began to ban its use in academic activities.
Some media have pointed out that certain responses generated by chatbots based on OpenAI technology “can be dangerous.” In this regard, the company is working closely with Microsoft on the integration of your technology in the Edge Internet browser. Microsoft has noted that user feedback is a crucial component in improving generative algorithms and a lot of work is being done to ensure that there is no chance of “causing” the AI ​​to generate unwanted responses.
ChatGPT works thanks to its training with large data sets of human-created content, but only after human moderators review these data sets according to the given guidelines on how to respond in different situations, especially when these scenarios are complicated.
In its existing version, ChatGPT has certain hard-coded safeguards against adult content and violent or hate speech. In these situations, the AI ​​platform should provide responses similar to “I can’t answer that.”
If the topic is considered controversial, ChatGPT tries to avoid bias by describing the views of people and movements, but does not take a very specific position.
Regarding the change presented, OpenAI explains that “Taking personalization to the extreme would risk enabling malicious uses of our technology and fawning AIs that mindlessly amplify people’s existing beliefs”.
Faced with that risk, the company ensures that there will always be limits to the behavior of the system. The challenge they must face is to define what those limits are. If ChatGPT attempts to make all of these decisions on its own, or if it were to be framed as a single, monolithic AI system, the company would claim that it would be in breach of its commitment to “avoid undue concentration of power,” set out in its letter of principles
