Not that I take pride in it, but the truth is that I’ve lost track of OpenAI’s releases. On Wednesday (16th), the company announced two new models, o3 and o4-mini, with intriguing developments.
The o3 is described by OpenAI as “our most powerful reasoning model”; the o4-mini is a “smaller model optimized for fast, cost-efficient reasoning.” Both are accessible through the ChatGPT UI and can handle various tools, such as “visual tasks” (analyzing uploaded image files).
One of the examples of visual task provided by OpenAI in the official announcement seems to have sparked a new craze: discovering the location of images based on the images themselves, a sort of reverse search or, as it’s been referred to on social media, “the end of Geoguesser.”
TechCrunch noted that the o3 isn’t much better than GPT-4o, an earlier and faster model, and that it’s not perfect, misidentifying the locations and sometimes failing to make a guess at all. Nonetheless, this capability of ChatGPT can be unsettling and already creates a new vector of paranoia regarding online privacy: it’s no longer enough to just clean the metadata from photos before uploading them.
By the very nature of LLMs, it’s challenging to distinguish genuine advancements from the enthusiasm of supporters. Techmeme, an aggregator of news and reactions from the tech industry, picked up this comment from someone on X:
I’m obsessed with o3. It’s way better than the previous models. It just helped me resolve a psychological/emotional problem I’ve been dealing with for years in like 3 back-and-forths (one that wasn’t socially acceptable to share, and those I shared it with didn’t/couldn’t help)
I find myself wondering what kind of “psychological/emotional problem I’ve been dealing with for years” a conversation with an AI released just hours ago could possibly resolve.
***
OpenAI’s blitz of releases is having an impact. In March, driven by trends like those from Studio Ghibli and the action figure boxes, ChatGPT became the most downloaded app in the world, according to consulting firm Appfigures, surpassing Instagram and TikTok, the usual leaders in recent months.
***
On Thursday (17th), Google released Gemini 2.5 Flash, which “delivers a major upgrade in reasoning capabilities, while still prioritizing speed and cost.” One of these days, a new model will be able to guess the color of our underwear and bring about world peace.