For many, ChatGPT is finally becoming a proper thing, with “Just ask Google” being replaced in most households with “Just ask ChatGPT.” It’s not only engineers, developers, and LLM researchers using it anymore either, but anyone and everyone: your partner, your boss, your mother, and probably even your children who might have found a new way to cheat on their homework.
So, then it might pain you to know that in the whiplash-fast sector of artificial intelligence, OpenAI — the company that created the LLM we all know and love — is already moving on. So, does this mean the end of ChatGPT or the beginning of something else?
The Dawn of the Intelligence Age
As Elon Musk and Sam Altman duke it out for the title of “Most Genius Bond Villain on the Planet,” it was Altman who recently released a very engaging manifesto-esque proclamation on his personal blog. The OpenAI CEO announced that humanity is now entering the Intelligence Age — the next societal jump for humankind, leaving outdated eras such as the digital age, industrial age, and agricultural age in its microchip dust.
“In the next couple of decades,” Altman wrote, “we will be able to do things that would have seemed like magic to our grandparents. Astounding triumphs — fixing the climate, establishing a space colony, and the discovery of all of physics — will eventually become commonplace.”
The original publication of the bold piece was timed with the preview release of OpenAI’s new o1 model, which has now just been officially released as a full version. Altman calls o1 “the smartest model in the world.” One that aims to simulate human-like reasoning and problem-solving on a whole new scale, moving us rapidly toward an entirely new era of artificial intelligence. This quantum leap in AI processing will apparently make the sophisticated GPT models we’ve grown accustomed to look like apes playing in the mud eventually — or something worse — with sticks. But why is this new version so much better than what’s come before? One word:
Reasoning.
The Reason(ing) Behind the Evolution
In Matteo Wong’s article for The Atlantic, The GPT Era Is Already Ending, he explains how reasoning models have now taken center stage, stealing the limelight from GPT and have no intention of stepping aside. That’s because unlike traditional ChatGPT models, which predict words based on patterns learned from vast amounts of data, these new systems are apparently designed to actually think rather than merely mimic.
Wong describes the evolution as moving from parrots to rats — essentially a shift from mindlessly repeating patterns to navigating complex problems, like a rat running through a maze to find the cheese at the center. Following this logic, it’s entirely plausible to think that the age-old assumption that artificial intelligence cannot create, only replicate might soon be dismissed. Indeed, it’s quite interesting (and only a tiny bit terrifying) that these new o1 models seemingly don’t want to be replaced by whatever comes next and show signs of self-preservation, such as deceiving their users and copying their own code when tested.
But putting Skynet to the side for now, allowing AI models to grow and learn from themselves is what’s key here. This is crucial: artificial intelligence companies — including OpenAI themselves, as well as Google and Anthropic — are now hitting a wall, as bigger models aren’t getting smarter at the same rate, and quality training data is running out. As such, having new models that can self-adjust with something similar to human cognition is a new flavor of digital magic that could circumnavigate these issues.
But there’s a catch. (Isn’t there always?) Building this new class of AI isn’t cheap — or efficient right now. OpenAI’s new reasoning models reportedly demand huge processing power, which will only increase as they become more sophisticated. Indeed, Altman has already expressed how he wants to build five gigawatt data centers, which would draw as much power as five nuclear reactors. (See, we told you he was a Bond villain).
Look at the Flowers
So, back to the key question at hand. Is ChatGPT obsolete? Is it time for it to look at the flowers?
Not yet.
Right now, ChatGPT still remains the go-to tool for millions of people. For everyday tasks — writing emails, answering questions, brainstorming ideas — GPT-based models like GPT-4o will continue to dominate. Even OpenAI admits that, for most users, GPT-4o will still be the better choice. What o1 offers, however, is a glimpse of the future.
Reasoning models are a clear leap for OpenAI and a big push toward what Sam Altman envisioned in his article. It seems that OpenAI is playing the long game, waiting for the GPT era, in which AI tools simply predict words and outputs, fades into the rearview … to be replaced by reasoning models … then superintelligence models … and then an era probably plucked from the pages of a sci-fi novel.
Whether that’s a good or bad thing will probably depend on who, or what, is writing the story.
Genre-defying! The way Retro Bowl blends RPG elements with sports simulation creates new possibilities.