GPT4, the artificial intelligence of OpenAi becomes even more powerful
Nothing else has been talked about in recent months: ChatGPT has ended up on websites, newspapers, TV news, bringing the concept of artificial intelligence to a large and non-specialized audience. And today, OpenAI, the company that invented the technology behind the chatbot, unveiled the new version, GPT-4.
“In 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,” OpenAI said. "GPT-4 is more reliable, creative, and able to handle much more complex instructions than GPT-3.5".GPT stands for Generative Pretrained Transformer, a reference to the fact that it is capable of generating text by itself and that it uses an artificial intelligence technology called transformers, which Google pioneered. It's a type of AI called a large language model (LLM), which is trained on large chunks of data collected from the Internet, learning mathematically to spot patterns and reproduce styles. Technically, the most important feature of the new version is multimodality, i.e. the ability to accept both text and image input and provide text output (the images will come later).
GPT-4 is also capable of generating much longer text strings and was designed to avoid the artificial intelligence pitfalls seen in the earlier GPT-3.5, OpenAI said. So much so that on the Bar Exam, GPT-4 ranks in the top 10% of scores compared to the bottom 10% of scores for the current version, GPT-3.5.
ChatGPT is more accurate. For example, it can instantly calculate a person's taxes and provide detailed image descriptions. But he can also write jokes and offer accurate medical advice.
OpenAI pointed out that the system underwent six months of security training and that in internal tests it was shown to be "82% less likely to respond to requests for disallowed content and 40% more likely to produce concrete answers with respect to GPT-3.5".
And yet, "in casual conversation, the distinction between GPT-3.5 and GPT-4 can be subtle. The difference emerges as the complexity of the request increases," OpenAI said. "GPT-4 is more reliable, creative, and capable of handling much more complex instructions than GPT-3.5."
GPT-4 also adds new "control" options. Users of large language models today often have to engage in complex prompt processing, learning how to insert specific cues into their prompts to get the right kind of responses. GPT-4 adds a command option that allows users to set a specific tone or style.
The new version of GPT also marks a step forward in avoiding a typical problem of artificial intelligence: hallucinations, or entirely made-up answers, often given with the same apparent authority as the correct answers. The hallucinations, however diminished, are still there, because artificial intelligence systems do not have an understanding of what is true and what is not, and therefore can generate completely false texts.