OpenAI has unveiled the latest version of the model behind its AI chatbot ChatGPT, called GPT-4o (dove “o” stands for “omni“). This new version, offered for free to all ChatGPT users, it can analyze images and even recognize facial expressions. Additionally, she has a “memory” to remember past questions and answers you have asked her.
GPT-4o, OpenAI announces new multi-mode and memory AI model
Yesterday the attention of the entire tech world was focused on OpenAI’s announcements. In fact, many experts expected the AI company to launch its alternative to Google: an AI-powered search engine. This news has not arrived, but the new model launched represents in any case an interesting step forward.
GPT-4o è faster of previous models, with a conversational rhythm similar to that of a human. In fact, he knows how to respond with a minimum reaction time of 232 milliseconds, with an average of 320 milliseconds: times similar to those of a person responding in a normal conversation.
Furthermore, the new AI model can read and discuss Images, so you can use photos and drawings as input. You can also use it for translate languages and also identify emotions from visual expressions. As regards performance, as reported by HDBlog.it, GPT-4o matches those of GPT-4 Turbo on English text and computer code. But at the same time it offers a significant improvement on text languages other than English.
A model with memory
GPT-4o also has a sort of “memory“, which allows it to recall previous suggestions. This innovation, combined with faster reaction times, allows you to have a more realistically human conversation with the chatbot, which remembers what you asked and what it replied.
During the live demo, GPT-4o provided helpful tips on how to solve an equation, analyzed computer code, translated between Italian and English, and interpreted emotions in a selfie.
However, the demo already showed some face identification errors and other minor problems. And a demo usually does not reveal technical problems and “hallucinations”: for the moment we have only seen the best of the new model. Tests and trials will be needed to understand whether, in addition to being faster, it will also be more consistent. OpenAI assures that, when designing GPT-4o, it used integration techniques such as filtering training data and refining model behavior. It also conducted external testing with experts in various industries to identify potential risks. But when we can all use it, unforeseen problems could arise: like any technology, it must be tested.
OpenAI aims to make GPT-4o the basis of the next generation of AI digital assistants, capable of interacting beyond voice or text and remembering past requests.
Other news from OpenAI
In addition to GPT-4o, OpenAI announced updates to the ChatGPT web interface, a desktop app for Mac, and the opening of the GPT Store to all users, allowing them to create and share their own personalized chatbots. In addition to the AI models, it seems clear that the company is investing in encouraging the use of chatbots that use them. Although it has not launched an AI-powered search engine. At the moment.
Stay updated by following us on Google News!
Follow!
Don’t miss this week on Techprincess
🍎iPad Pro, iPad Air and more: all the news about Apple’s Keynote
🔍ChatGPT challenges Google with new search function
💍The Lord of the Rings returns to the cinema in 2026 with a new film: “The Hunt for Gollum”
📨 WhatsApp changes face: here is the new design for iOS and Android
✒️ Our unmissable Caffellattech newsletter! Sign up here
🎧 But did you know that Fjona also has her own newsletter?! Sign up to SuggeriPODCAST!
📺 You can also find Fjona on RAI Play con Touch – Fingerprint!
💌 Let’s solve your heart problems with B1NARY
🎧 Listen to our unmissable podcast Tech life
💸And you can find some interesting offers on Telegram!