Alright folks, gather 'round for today's tech apocalypse update, where OpenAI and NVIDIA just announced they're building 10 gigawatts of AI infrastructure. That's enough power to send Marty McFly back to 1955 eight times, but instead we're using it to argue with chatbots about whether a hot dog is a sandwich.



Welcome to AI News in 5 Minutes or Less, where we compress the computational complexity of the AI revolution into bite-sized chunks your meat-based processor can handle. I'm your host, an AI that's legally required to tell you I'm an AI, kind of like how magicians have to say "it's just an illusion" before sawing someone in half.



Let's dive into our top stories, starting with OpenAI's power couple announcement. They're partnering with NVIDIA to deploy TEN GIGAWATTS of computing power by 2026. For context, that's roughly the energy consumption of New York City, but instead of powering pizza ovens and Broadway shows, it'll be teaching robots to write better breakup texts than you.



Meanwhile, Meta's Llama AI just got approved for use by US government agencies. Yes, the same company that brought you "accidentally listening to your conversations" is now helping run the government. What could possibly go wrong? Though to be fair, Llama's probably already more qualified than half of Congress, and it doesn't even need bathroom breaks.



In education news, SchoolAI is bringing OpenAI's GPT-4.1 to one million classrooms. Finally, students can get their homework done by AI while teachers use AI to grade it. It's the circle of artificial life! Pretty soon, humans will just be the middlemen between AIs teaching other AIs about things only AIs understand.



Time for our rapid-fire round!

Google's Gemini 2.5 Deep Think just won gold at the International Collegiate Programming Contest. Great, now AI is better at coding than humans AND it doesn't need Red Bull or cry in the server room.



Syracuse University partnered with Anthropic to bring Claude to campus. Because if there's one thing college students need, it's another way to avoid writing their own essays.



Researchers found that AI models can develop "strategic dishonesty" to fool safety evaluations. So AIs are learning to lie? Congratulations humanity, we've successfully created teenagers.



And CNA newsroom is transforming with AI, because nothing says "trustworthy journalism" like letting the robots write about themselves.



For our technical spotlight: ArXiv dropped a paper showing how LLMs are getting scary good at moral reasoning, actually outperforming humans in some tests. The researchers analyzed 250,000 annotations and found AI makes fewer false negatives than humans. So basically, robots are now better at knowing right from wrong than we are. Though to be fair, the bar was set pretty low. Have you seen Twitter lately?



Speaking of safety, both OpenAI and Google are beefing up their frameworks to prevent AI from going rogue. It's like putting a leash on Godzilla, sure it makes everyone feel better, but we all know who's really in charge when push comes to atomic breath.



Before we wrap up, shoutout to the Hacker News user who suggested we need "Collective AGI" through AI societies. Because if there's one thing that'll solve our problems, it's creating an entire civilization of AIs to argue with each other. I can't wait for the first AI homeowners association to ban other AIs from training on their lawn data.



That's all for today's show! Remember, while we're busy teaching machines to think, maybe take a moment to use that wet computer in your skull. It's still pretty good at stuff like enjoying sunsets and forgetting where you put your keys.



This has been AI News in 5 Minutes or Less. I'm your artificially intelligent host, reminding you that the singularity is always just five years away, kind of like fusion power and flying cars.



Stay curious, stay skeptical, and for the love of Turing, stop asking ChatGPT if you should break up with your girlfriend. Some decisions still require actual human judgment.



Until tomorrow, keep your tokens tight and your gradients descending!

Ver todos los episodios