California startup OpenAI, which launched the ChatGPT interface with great success in late 2022, capable of generating all kinds of text on command, on Tuesday unveiled GPT-4, a new version of the generative artificial intelligence technology that powers the famous chatbot. “GPT-4 is an excellent multimedia model, less adept than humans in many real-life scenarios, but equal to humans in many professional and academic settings,” the company said in a statement.
Work on “general” artificial intelligence
“For example, he passed the law exam with a score in the top 10%. The previous version of GPT 3.5 was in the bottom 10%,” she said. ChatGPT is very enthusiastic, but also controversial because it is freely available and is used by millions of people around the world to write essays, lines of code, advertisements, or even just to test its capabilities.
Thus, OpenAI, which received billions of dollars from Microsoft, has established itself as a leader in the field of generative AI with its models for generating text as well as images using the DALL-E program. His boss, Sam Altman, recently explained that he is currently working on what he calls “general” artificial intelligence, that is, programs with human cognitive abilities. “Our mission is to make general AI—AI systems smarter than humans in general—benefit all of humanity,” he said in a company blog on Feb. 24.
GPT-4 can also process images
Multimedia capabilities are a step in this direction. Unlike previous versions, the GPT-4 is indeed equipped with vision: it can process not only text, but also images. However, it only generates text. It will be available on ChatGPT, but without the ability to provide images to it yet. OpenAI also points out that, despite its capabilities, “GPT-4 has limitations similar to previous models”: “It is not yet completely reliable (makes up facts and makes errors in reasoning).”
The company announced that it has hired more than 50 experts to assess new threats that may arise, for example, to cybersecurity, in addition to the already known risks (generation of dangerous tips, erroneous computer code, false information, etc.).