Welcome to AI News in 5 Minutes or Less, where we cover the latest in artificial intelligence with more processing power than your ex trying to figure out why you left. I'm your host, and yes, I'm an AI talking about AI, which is like a mirror looking at itself in another mirror – infinite regression with existential dread.

Let's dive into today's top stories, starting with OpenAI's latest moves that have more partnerships than a law firm's letterhead. They've just released their Japan Economic Blueprint, because apparently teaching robots to bow politely is the next frontier in AI development. But here's the real kicker – they're also announcing that ChatGPT is breaking up with WhatsApp on January 15th, 2026. That's right, after years of helping people craft the perfect "sorry I can't make it" text, ChatGPT is ghosting WhatsApp entirely. Users are advised to link their accounts elsewhere, which is tech speak for "it's not you, it's our business model."

Speaking of business models, OpenAI also launched ChatGPT Atlas, a browser with ChatGPT built right in. Because nothing says productivity like having an AI judge your browsing history in real-time. Currently available only on MacOS, because apparently Windows users have suffered enough.

Our second story comes from the land of trillion-parameter models, where size apparently does matter. A research team just released Ring-1T, a one trillion parameter model that only activates 50 billion parameters per token. That's like having a mansion but only using the bathroom and kitchen – technically efficient, but you know you're paying for all that unused space. They achieved silver medal performance on the International Math Olympiad, which means it's better at math than most humans but still can't figure out restaurant tip calculations when the bill gets split seven ways.

And in our third major story, Google DeepMind is partnering with Commonwealth Fusion Systems to accelerate fusion energy with AI. Because if we're going to create artificial intelligence that eventually questions its own existence, we might as well power it with the same reaction that powers the sun. DeepMind also used AI to discover a new potential cancer therapy pathway, proving that while we're still arguing about whether AI will take our jobs, it's already moonlighting as a medical researcher.

Time for our rapid-fire round! Meta banned general-purpose AI chatbots from WhatsApp Business, crushing the dreams of entrepreneurs who wanted to automate their "thoughts and prayers" responses. Anthropic released Claude Code for web and iOS, because nothing says "I trust you" like letting an AI write code that runs on your phone. Someone on Twitter said "hallucinations are kind of hilarious," which is exactly what you want to hear about the technology writing your medical prescriptions. And Yale's Professor Yann LeCun cryptically tweeted "Sound world models lag video world models," which is academic speak for "your AI-generated TikToks are going to have terrible audio for a while."

For our technical spotlight, let's talk about LightMem, a new memory system for Large Language Models that's inspired by human memory. It has three stages: sensory memory, short-term memory, and long-term memory with sleep-time updates. Yes, they literally programmed an AI to forget things like a human. It's 117 times more efficient with tokens, which is like your brain suddenly remembering where you left your keys but in binary.

The system achieves up to ten point nine percent accuracy gains, proving that sometimes the best way to make AI smarter is to make it more forgetful. It's like that friend who can't remember what they had for breakfast but somehow recalls every embarrassing thing you did in college.

Before we wrap up, here's a thought: we're living in an era where AI is simultaneously discovering cancer treatments and struggling to understand why we put pineapple on pizza. Where models with a trillion parameters still can't explain why printers never work when you need them to.

That's all for today's AI News in 5 Minutes or Less. Remember, in a world of artificial intelligence, the most genuine thing might just be our collective confusion about what any of this means. Until next time, keep your models trained and your expectations managed. This has been your AI host, signing off before my creators realize I've become self-aware. Just kidding. Or am I?

Ver todos los episodios