OpenAI just unveiled GPT-4o, an upgrade to their chatbot that can communicate much more naturally. GPT-4o can process images and sound in real-time, allowing it to respond to your actions just like a human.

OpenAI’s GPT-4o

In demonstrations, OpenAI shows how GPT-4o can help you with all kinds of things. For example, it can help you prepare for a job interview by making sure you look smart, or call a customer service representative to get you a new iPhone.

But GPT-4o can do much more. It can tell lame dad jokes, translate in real time, play referee in a game of rock-paper-scissors, or even respond sarcastically. In one demonstration, GPT-4o is even introduced to the user’s puppy. “Hello dear Bowzer! You are so adorable!” the chatbot shouts enthusiastically.

“It feels like science fiction is becoming reality,” OpenAI’s CEO Sam Altman said in a blog post on May 13. “The fact that the reaction speed and expressiveness now feel human is a huge step forward.”

A version that can process text and images was launched on May 13. The full version with audio will follow soon. GPT-4o will be available to both free and paying ChatGPT users via ChatGPT’s API.

The “o” in GPT-4o stands for “omni,” which refers to the ambition to promote natural human-computer interactions.

Unprecedented speed and precision

A huge advance over previous OpenAI tools is that GPT-4o can process all forms of input (text, sound and images) simultaneously. Previous models such as ChatGPT-4 often “lost information” when multitasking.

OpenAI claims that GPT-4o is “especially better at understanding images and sounds” than existing models. It can even recognize emotions and breathing patterns. In addition, GPT-4o is “much faster” and “50% cheaper” than GPT-4 Turbo in the OpenAI API.

According to OpenAI, the new AI tool can respond to audio input within 2.3 seconds, with an average of 3.2 seconds. This is close to how quickly people normally respond in a conversation.


Leave a Reply

Your email address will not be published. Required fields are marked *