ChatGPT was in the spotlight at the beginning of the year, but one feature that worries users is how this artificial intelligence (AI)-generated content can be offensive in certain cases and Open the AI They talked in detail about how they are going to handle this situation.

Via the official blog, the company revealed that it is developing an improved version of the artificial intelligence to be able to “customize” it.

This is the result of certain results shared by users who “find them to be politically biased, offensive, or reprehensible.”

This may interest you: Windows adds artificial intelligence to Bing search engine to create images with text

Custom GPT chat

is artificial intellect It works as follows: it has a series of neural networks that are first pre-trained thanks to a large data set, then move to a phase in which the model is polished by a series of reviewers. which regulate them according to certain guidelines.

This endpoint is something the company wants to improve. from Open the AI They acknowledge that the process is imperfect and ensure that “we will provide clearer guidance to reviewers about potential issues and challenges related to bias, as well as controversial figures and issues.”

You may also be interested in: ChatGPT: How to Create Contracts, Essays and Emails Using Artificial Intelligence

They also note that they will attempt to enhance demographic bias and use enhancements based on techniques such as Constitutional AIin which people’s work is focused only on setting learning rules.

“Our instructions are clear that reviewers should not weigh in on any political group. The biases that may arise from the process described above are errors, not features,” the organization said.

At a general level, the company will work on three pillars to solve the problem of content that may be offensive or has a strong ideological bias:

– Improve default behavior: The goal is to eliminate bias and also cases where the system “makes things up” because it affects the position of users.

– Personalization: ChatGPT allows users to modify their behavior by allowing responses to be generated that they disagree with. Although the limits will be set in the same way, so the challenge that they consider in the company is to find a balance in these barriers to avoid the “misuse” of the technology.

– popular participation: They want people to “influence the rules of the system” and include “as many perspectives as possible” that are relevant to educational, social, political and other fields.

All of this learning will be a work in progress, so the company said it will provide more details “over the next few months,” and they hope to continue monitoring users for more bug fixes.

You might be wondering why AIs like ChatGPT would be a problem for the environment

Microsoft will integrate ChatGPT into iPhones

then Microsoft Announced the incorporation of artificial intelligence ChatGPT Your search engine system, Bing, a company technology New ways to access this platform through additional devices have begun to be developed, as it is currently only available in a limited way through it web version or in the application limit.

One of the first ways in which Microsoft Trying to enter the mobile application market with this search engine is Bing by creating a platform specifically for devices that use the operating system. iOS.

Although expanding the scope of the updated Bi search engineNog will start with the desktop platform and then on devices like iPhoneThis does not mean that the users of the device Android They will not benefit. It is likely that the development of a native application that combines these capabilities with this operating system in mobile phones will soon reach market.

According to the CEO of the company Microsoftsatya nadella artificial intellect which is integrated into the Bing service, will allow people to perform the desired search from a browser or a special application for mobile devices. is characteristic Makes the customer feel like their process navigation “Copilot” was in charge.