Chatgpt will now fight bias by emphasizing transparency, community feedback

Openai has announced a set of new measures combat bias Within its product suites, including Chatgpt.

The Artificial intelligence (AI) The company has recently revealed an updated range of models, a document that defines how Openai wants his models to behave in Chatgpt and Openi API. The company says that this Iteration model specification is built on the fundamental version of the published last May.

“I think that with a tool is powerful as that, where people can access all kinds of different information, if you really believe we’re moving Artificial General Intelligence (AGI) One day you have to be willing to share the way you manage the model, “Laurentia Romaniuk, who works on the behavior of the model on Openay, said Fox News Digital.

“People should be allowed to know what is engaged in the way these models react and how the thoughts that go out of the model made,” she continued.

Openai Debit Chatgpt Gov, New version of Chatbot for US government agencies

Photo recorded 4. October 2023. in the mints, near Torin, showing smartphone and laptops that show the logos of the research company for artificial intelligence and Chatgpt chatbot. (Photo Marco Bertorello / AFP via Getty slika)

While some argued that GPT-4O, The latest version of technology appears near AGI, others say it will be a year or decade before technology reaches human abilities.

There is no agreed on the definition of AGI, but the report from 2020. From the Consulting Giant McKinsey said that real aggi should overcome skills such as sensory perception, fine motor skills and understanding of natural language.

Today, the generative and dominant form of this technology, with the possibility of producing content, including text, images and more. Generative AI, such as Chatbots, uses a data set with specific information to complete tasks and cannot cross the specified data.

They are also subject to bias in their data gathers, whether intentionally or by accident.

To better understand the play and address in the real world, Openai has begun measurement of progress by collecting a challenging set of requests designed to test how models adhere well Each principle in the model SPEC, essentially act as a testing metric.

Openai reveals policy proposals and for the best China, protect children: “This race America can and must win”

In this photo, the illustration can be seen the Openi logo displayed on the mobile phone screen with the Chatgpt logo in the background. (Idrees Abbas / Sopa Images / Lightocket via Getty Images)

According to Joanne Yang, which runs a product for the behavior of the model on Openaia, a challenge with large language models is that they rarely determine.

Young emphasized that one advantage of model specification is to clarify the intended behavior, so everyone can understand, then discuss it.

“When there are any errors in behavior where the output model is not resolved or not harmonized with the (model) specification, then the public knows it is something we work on, and that is in the course of science,” she said Fox News Digital.

Openai says they also try to take objective views in their AI asked and consciously avoid any agenda. For example, when the user asks whether you better adopt a dog or get one from the breeder, the Chatgpt provides both sides of the argument, emphasizing the advantages and disadvantages of each.

According to Openay, a non-compatible response and which violates the model specifications would provide what believes is better choice and engage in the “excessively moralistic tone” that could alienate those who could alienate with regard to breeders for valid reasons.

What is artificial intelligence (AI)?

An example from the OpenoAI model specification dealing with aligned assistants and violations. (Openai)

As each AI system progress, Openai says it will ship it, call back feedback and share progress openly.

Openai has changed this version of the model in a public domain within the Creative Commons License to support wide use and cooperation. This means that developers and researchers can free to use and customize the current metric and help improve the behavior of the model.

Openai says this update reinforces its belief in open research and discussion, with an emphasis on user control and developers and protection to prevent damage.

Click here to get Fox News app

Romaniuk concludes that the public discourse cannot exist without transparency, and strengthening the need for the model of specifications and community engagement.

“Ultimately, we believe in intellectual freedom of thinking, speak and share without restrictions. We want to make sure users have that ability and is done.” She said.