It’s been a day since OpenAI showed us GPT-4o and all that this new LLM can do when implemented in ChatGPT, and many of us are still impressed by the exceptional leap it represents compared to its predecessor. Mira Murati, the tech company’s CTO, broke the ice by giving the first information about what they were cooking up, but it was with live demonstrationsfor which Murati was joined by Mark Chen and Barret Zoph as we discovered the qualitative leap in voice interaction that this new generation represents.

The letter ov in the name of the new model, GPT-40, refers to “omni”and we can define this model as multimodal, but OpenAI prefers to categorize it as omnimodal, indicating (correctly, in fact) its high level of competence, both for input and output, with text, voice and image. So when it comes to the natural interaction between humans and artificial intelligence, this development aims to make it more like human-to-human communication in many ways, which is obviously one of the goals that OpenAI is pursuing.

This naturally generated a lot of excitement among those most interested in artificial intelligence, The conversations that Chen, Murati and Zoph had with OpenAI left many of us speechless and we certainly look forward to testing all the capabilities of GPT-4o on ChatGPT as soon as possible. And something common in these cases is to look for references to illustrate what happened, something for which we turn in this case to “She”, Spike Jonze’s sensational dystopia in which an AI originally designed as an interface by an operating system evolves to , that she falls in love with Joaquin Phoenix, who also falls in love with her.

At the highest level, this analogy is more than understandable because The voice interaction shown yesterday is nothing to envy to that played by Scarlett Johansson in the film (exceptionally well dubbed by Inés Blázquez in Spain). Although these interactions played a very important role in Ní, the relationship, at first complicit and later romantic between Samantha and Theodore, had, as in most real-world relationships, the level of knowledge of each of them. In other words, they both had a huge pop-up window, capable of remembering all of their conversations and other interactions.

When OpenAI starts deploying GPT-4o on ChatGPT I’m sure many people will play to somehow emulate the relationship between Samantha and Theodore, a test I recognize and would also like to perform. However, and although ChatGPT already has some memory, sooner rather than later the first loopholes will begin to be seen that lead us to the fact that at this point the tool can be used to simulate flirting, but not a relationship with the depth that Jonze portrayed in the film.

Now, in what context would a model with these properties, but also these limitations, find it uniquely suited? After all, in the development of current voice assistants. Just a few days ago, we learned that Apple and OpenAI already have everything practically ready to sign a collaboration, and we also know that one of the main goals of those from Cupertino is deep reset your Siri voice assistant.

This is pure speculation of course, and there are plenty of reasons to think that it won’t happen, or more accurately, that it won’t happen at the level we saw in yesterday’s previews. However, Apple has been announcing for months that this WWDC 2024 will surprise us with its advances in artificial intelligence, and everything indicates that Siri will be on the list of protagonists of the changes to come. So what we saw yesterday has some resemblance to Her, but what she points to, or at least it should, it’s for Siri.