ChatGPT-4: an artificial intelligence capable of conversing like a human, provokes fear and interest.
Created on 21 March, 2023 • 1 minutes read
On Tuesday, OpenAI announced the creation of a multimodal artificial intelligence model, GPT-4, capable of processing input data in the form of text and images and producing text
On Tuesday, OpenAI announced the creation of a multimodal artificial intelligence model, GPT-4, capable of processing input data in the form of text and images and producing text output with performance comparable to that of a human in various professional and academic tests. On the same day, Microsoft announced that its Bing Chat works on top of GPT-4.
If GPT-4 functions at the level claimed by the developers, it could become a new era in the development of artificial intelligence. According to OpenAI, the GPT-4 model successfully passed a bachelor's degree-level simulated exam with results better than those of 90% of graduates, while the previous GPT-3.5 model showed worse results than those of 90% of graduates.
Currently, access to GPT-4 is provided to ChatGPT Plus subscribers. Additionally, the company is testing the capabilities of GPT-4 in generating images using the Be My Eyes smartphone app, which can recognize scenes and describe them.
The multimodal capabilities of GPT-4 include analyzing the content of multiple images, understanding jokes using multiple sequences of images, and extracting information from a diagram. Microsoft and Google are also experimenting with multimodal capabilities. Microsoft believes that a multimodal approach is necessary to achieve "general artificial intelligence" or "strong AI" that can perform common tasks at a human level.