OpenAI revolutionizes artificial intelligence with GPT-4o: what it is, what the “o” means and whether it will be free or not – La Linterna
Last week, Mira Murati, CTO of OpenAI, introduced a new artificial intelligence model: GPT-4o. A model that promises to be a leap in the evolution of the popular GPT chat: much faster, smarter and more versatile.
The presentation was so impressive that it left no one indifferent: for some it was impressive progress, for others it was an unprecedented threat. What is this new GPT-4o and how does it work?
Photo of an illustration showing the ChatGPT page view in Beijing (China) / EFE.
What is GPT-4o and what does “o” mean?
First of all: what is GPT-4o? Technology communicator Mario Yanez believes that the evolution of this new model is that it is able to interact with humans in a much more natural way. Thus, this new version is “multimodal”, that is, “can receive inputs/questions or information in text, audio or video mode, hence the “o” in the name: “omnimodal”, explains the La Linterna participant.
Plus, he explains, it’s much faster and can respond in less than 0.3 seconds on average. “This will allow us to use this new version as a real-time virtual assistant, like Siri or Alexa, but much smarter.”
The fact is that those who were able to see the demonstration that OpenAI engineers made said that it was something impressive. They note that the most impressive thing was his ability to conduct a lively conversation: “he can carry on a conversation exclusively with his voice, he can be interrupted, he knows how to stop, listen to us and continue, even changing arguments.”
Here, OpenAI shows how GPT-4o can translate from Italian to English in real time. The world is beginning to experience a huge transformation with the advent of AI.
– Fernando J. Valls??? (@FernandoHWalls) May 13, 2024
- Left6: There is no advertising configuration for the requested slot.
But the most impressive thing is that it can change tone: “At one point the model was reading the text and was asked to speak more dramatically, and the model became more and more theatrical until Mira Murati herself told him.” He asked me to switch to a convincing robotic voice, and he did it perfectly,” says Yanez.
Will GPT-4o be free?
Another example that was presented at the GPT-4o presentation was where one of the engineers recorded himself writing an algebraic equation (like 3x+1=4) on a piece of paper, and the GPT-4o followed suit. So they asked him to guide them as a teacher guides a student. And when he set up the model, he replied, “The first step is to get all the terms with x on one side of the equation,” “What do you think we should do with this plus one?”
Multimodal model GPT-4o/EP
However, the popularizer of the technology warns that it is still a mathematical model of language “with its limitations.” In fact, during the demonstration, the voice of GPT-4o began to comment on the outfit of one of the presenters, “even though he was not asked to do so.” And, like previous models, it may have “bloopers”.
But will GPT-4o be free? Many of OpenAI’s most powerful features, such as reasoning with images and videos, were already available, but they came at a cost, Yanez said. “What has been done with GPT-4o is to combine these text, video and audio capabilities (multimodality) into a single robot and open up these functions to the general public and something else for free,” he emphasizes. Of course, the difference with the paid version will be that paid users will have 5 times more features.
Mario Yañez, technology distributor and La Linterna employee
Another feature of the payment mode will be the ability to remember old requests if we voluntarily activate this feature. “This will allow the model to better contextualize a new request or question based on previous conversations. You will be able to get to know us better and tailor the answers to our way of expressing ourselves or to the topics that interest us. Even in the future, it will be able to interpret our facial expressions and know whether we are sad, happy or angry.”