Start-up OpenAI, which at the end of November successfully launched a ChatGPT interface capable of generating various types of text, revealed on Tuesday the new GPT-4 version of the generative AI technology that powers the popular chatbot.
The company said, in a statement, that the “GPT-4 version is a remarkable multimedia model that is less adept than humans in many real-life scenarios, but just as good as humans at performing many professional and academic tasks.”
The company added that the new version of the performance can “for example, pass the bar exam with a score as good as the top 10% of those who passed. The previous version, GPT 3.5, was below 10%.”
The ChatGPT algorithm arouses a lot of excitement and controversy also because it is free to access and millions of people around the world use it to write articles, lines of code, advertisements or even simply to test their abilities.
OpenAI, which has received billions of dollars from Microsoft, has proven to be a leader in generative AI through text-as well as image-generating models through its DALL-E software.
Sam Altman, the company’s president, said recently that work is underway on so-called “general” artificial intelligence, that is, software with human cognitive capabilities.
“Our mission is to ensure that general AI systems, systems that are generally smarter than humans, are beneficial to all of humanity,” he said in the company’s blog on February 24.
Multimedia capabilities are a step in that direction.
Unlike previous versions, the GPT-4 version can handle text and images as well. However, it currently only generates text.
The version will be available on ChatGPT but without the ability to provide it with images at this time.
#version #Chat #GBT #advanced