It was on everyone's lips, there was no talk of anything else. Today, OpenAI, the company that invented artificial intelligence technology, unveiled the new version. OpenAI launches GPT-4, the powerful new model for understanding images and text that the company calls "the latest milestone in its effort to scale deep learning." "In a casual conversation, the distinction between GPT-3.5 and GPT-4 can be subtle. The difference emerges when the complexity of the task reaches a sufficient threshold," OpenAl said. Multimedia, this is the most important feature of the new version, that is, the ability to accept both text and image inputs and to provide text output, for the moment the images will come later. The most innovative feature of GPT-4 is its multimode input functionality. However, although AI is able to process data from different sources, such as photos, images and graphs, the output generated will always be in text format.
So GPT-4 is able to generate text and accept image and text inputs , an important novelty compared to the last version that accepted only text and achieves "human level" performance in various professional and academic benchmarks. The technology used is NLP (natural language processing) that allows Chat GPT to study and understand the patterns and even nuances of human language, and learn thanks to trained machine learning mechanisms, based on a huge amount of data.
Users will be able to leverage GPT-4 to analyze a wide range of materials, from documents containing graphs and figures, to textbooks, photos and screenshots. GPT-4 is able to understand and synthesize all this information, reformulating it in different ways and tones, customizable according to the needs of the user. Despite its improvements, GPT-4 still has some imperfections and limitations typical of previous language models. A known problem, in the previous version, is the tendency to generate invented and wrong answers. This is due to the lack of understanding of the true and false by artificial intelligence systems.
However, compared to GPT-3, GPT-4 showed considerable progress. In addition, the likelihood of GPT-4 generating violent or harmful text in response to disallowed content requests decreased by 82%, making the system safer in everyday use. Improvements also compared to sending more words than the previous version, GPT-4 will allow 25,000 words to be sent, allowing the user to analyze a larger amount of material without having to break up the work. GPT-4 available in Italy through ChatGPT Plus, a paid service with monthly subscription. In conclusion, GPT-4 represents a significant step forward in generative artificial intelligence technology, offering new opportunities and potentials for companies, developers and end users.