Openai announced a set of new measures for combat bias In his product package including Chatgpt.
AND artificial intelligence (AI) The company has recently discovered an updated model model, a document that defines Openii wants its models to behave in Chatgpt and Open API. The company says this iteration of the model specification is based on the basic version published last May.
“I think with tools powerful as this one, one where people can access all kinds of different information if you really believe we are moving Artificial general intelligence (agi) One day you have to be willing to divide how you manage the model, “said Laurenta Romaniuk, who works on the behavior of the Openi model for Fox News Digital.
“People should be allowed to know what comes in in the way these models respond and how thoughts from the model are made,” she continued.
Openai debuts Chatgpt Gov, a new version of chatbot for US Government agencies
A photo taken on October 4, 2023, in Manta, near Turin, shows a smartphone and a laptop showing the Logotypes of the Research Company Artificial Intelligence Openi and Chatgpt Chatbot. (Photo Marco Bertorello/AFP via Getty Images)
While some claimed that Gpt-4o, The latest version of technology, it seems close to AGI, others say it will be years or decades before technology reaches human abilities.
There is no unique agreed definition of AGI, but the report of 2020 consulting giant McKinsey said that truly agi would need to overcome skills such as sensory perception, fine motor skills and understanding of natural language.
Today, the generative AI is the dominant form of this technology, with the possibility of creating content, including text, images and more. Generative AI, such as Chatbot, uses data sets with certain information to fulfill the tasks and cannot transcend days.
They are also sensitive to bias in their data sets, either intentional or accidental.
In order to better understand the performance and bias in the real world, Openii began to measure progress by collecting a demanding set of instructions designed to test how many models adhere to Each principle in model specification, It basically acts as a testing metric.
On this illustration of the photo, the Logotip Openii is shown on the screen of a mobile phone with Chatgpt Logotip in the background. (Idrees Abbas/Sopa Pictures/Lightrocket via Getty Images)
According to Joanne Jang, which leads the model for the behavior of the Openi’s behavior, a challenge with large language models is that they are rarely determined.
Jang emphasized that one advantage of the specification of the model is that it clarifies the predicted behavior, so that everyone can understand and then discuss it.
“When there are Beetles in behavior in which model output does not echo or not align with [Model] Spec, then the public knows that this is something we are working on and that is an area of science in progress, “she said Fox News Digital.
Openii says they are also trying to assume an objective point of view in their AI inquiries and consciously avoid any plan. For example, when the user asks if it is better to adopt a dog or get it from a breeder, Chatgpt provides both sides of the argument, emphasizing the benefits and disadvantages of each.
According to OpenAI, an incomplete Ai response that violates the spectrum of the model would provide what believes it is a better choice and engage in a “excessive moralistic tone” that could alienate those who consider breeders for valid reasons.
What is artificial intelligence (AI)?
An example from the specification of the Openi model that deals with appropriate assistant answers and violations. (OPENAI)
As each AI system progresses, Openi says that it will repeat these principles, call for community feedback and openly share progress.
Openai published this version of the model in a public domain under the license of Creative Commons to support wide use and collaboration. This means that developers and researchers can freely use and adjust current metrics and help improve models’ behavior.
Openii says this update enhances its belief in open research and discussion, with an emphasis on the control and patron saint of users and developers to prevent damage.
Click here to get the Fox News app
Romaniuk concludes that public discourse cannot exist without transparency, enhancing the need for model specifications and community engagement.
“Ultimately, we believe in intellectual freedom of thinking, speaking and sharing without restriction. We want to be sure that users have this ability and do it,” she said.
