ChatGPT’s underlying language model, GPT-3.5, is about to be replaced. Speaking to the audience at the company’s “AI in Focus” event, his CTO of Microsoft Germany told the audience that GPT-4 will be released soon, unlocking new features, including video.
ChatGPT has sent shockwaves through most of the world. The fastest growing app in history, this free chatbot sandbox marks the dawn of a new era in which neural networks can communicate almost as persuasively as humans do. He single-handedly alerted the public to the fact that It’s also convenient for writing code. It is far from perfect and often very wrong, but its rise portends nothing less than a fundamental upheaval in human economics and social fabric. Also, it’s a lot of fun to play.
ChatGPT is built on the brains of OpenAI’s Generative Pre-Trained Transformer (GPT 3.5 language model). Essentially, GPT captured an unprecedented amount of human writes. Billions of web pages, billions of books, billions of code snippets, millions of human conversations. We learned how to analyze this treasure trove of information and write like we do. Ask a question or give it a task and within seconds you’ll get the kind of answer you’d expect this type of question to typically receive.
Details are currently not my forte. Its responses often display a surprising degree of contextual understanding and insight, with well-structured arguments and very naturally readable text, but much of its output is not factually correct, so the truth Sex is absolutely unreliable. extreme confidence.
Well, according to hot online, this amazing brain has received quite an upgrade. “Next week we will introduce GPT-4,” said Andreas Braun, CTO of Microsoft Germany at last Thursday’s AI in Focus event, adding that “multiple applications offer completely different possibilities, such as video. We plan to have a modal model.”
This multimodal approach allows GPT to learn not only from text, but also from other media such as audio and video, opening up a huge new hodgepodge of information available to the system.
Exactly what the outcome will be is unknown. Training his GPT 3.5 with hundreds of billions of bits written is already a huge processing task, and text is a very dense form of information. Opening the door to audio and video appears to greatly increase the time and processing power required to capture and analyze information. Similarly, if GPT initiates responses in audio or video format, it is unlikely that OpenAI will consume its processing and bandwidth costs.
But we’ll find out soon enough. Again, this is just the tip of the spear. Neural networks will be able to ingest and output information the same way humans can. Of course, GPT has to learn to understand audio and video. It will be interesting to see how long it takes for these ingenious and enigmatic networks to be able to have real-time conversations.
sauce: hot online
