Shulou(Shulou.com)11/24 Report--

On the evening of October 18, Beijing time, OpenAI said it was developing a new tool that could determine with considerable precision whether an image was drawn by artificial intelligence.

In addition to chat robots and writing tool ChatGPT, OpenAI has also developed an artificial intelligence image generator called DALL-E.

Mira Murati, the company's chief technology officer, revealed on Tuesday local time that the company's AI image recognition tool can determine with 99 per cent accuracy whether an image was drawn by an artificial intelligence tool.

Murati said the image recognition tool is under internal testing and will be released later, but she did not provide a specific release date.

OpenAI's ChatGPT has brought about a revolution in generative artificial intelligence technology. Many products on the market can generate pictures based on keywords entered by users. On the other hand, some artificial intelligence companies have also launched recognition tools to determine whether certain pictures or articles are created by artificial intelligence tools, but the accuracy of these tools is not high.

Earlier this year, OpenAI also launched an identification tool to identify whether an article was created by an artificial intelligence assistant, but unexpectedly, the product was removed from shelves in July because it was not accurate and unreliable for users.

When the shelves were removed, OpenAI said it would continue to improve the recognition product, as well as identify content created by other artificial intelligence assistants, such as pictures, audio, etc.

Although generative artificial intelligence brings convenience to many people, it also brings other worries. For example, in the news field, when major events occur, some bad people will use artificial intelligence tools to make up news articles and spread them around. This has given rise to the need for identification tools.

The creation of artificial intelligence also brings another problem, which is to damage the copyright of other works. Recently, the Firefly picture generator team of Adobe announced that it would never let pictures created by artificial intelligence infringe upon other people's copyright.

OpenAI's big language model and ChatGPT are industry stars, and on Tuesday, company executives also talked about plans for a follow-up version of the GPT-4 big language model.

So far, OpenAI has not announced the name of the follow-up big language model, but data show that in July this year, the company applied for a new trademark, "GPT-5", with the United States Patent and Trademark Office.

As many users have found, ChatGPT, an artificial intelligence assistant based on GPT-4 and GPT-3.5 large language models, will get the wrong answer and fall into the so-called "artificial intelligence illusion". Murati's answer to the media question "whether the GPT-5 language model will continue to give the wrong answer" is "possible".

Murati said that the development team has made some progress on the artificial intelligence illusion of the GPT-4 model, but it is clear that the company has not reached the level it should be.

It is reported that artificial intelligence manufacturers need to use high-performance processors to train large language models, and Nvidia is currently a leader in this processor market. On Tuesday, Sam Altman also talked about whether OpenAI will develop its own processor chips for training.

"by default, we certainly won't develop chips, but I don't want to rule it out," Altman said. "

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.