This Week in AI: The Next Llama is Running Late

Artificial intelligence can already do so much, from producing a song or creating an image from a simple line of text, to building an app. It can even rewrite your correspondence in the voice of Yoda from Star Wars, as if he were in a Shakespearean play.

But what comes next, you might wonder? Or said another way: "A spectacle most grand, those powers it has. But this? The next iteration, much wanting, it will leave you, hmmmm? Quite unworthy."

AI companies reportedly are struggling to demonstrate impressive upgrades for their next generation technologies. The upgrade from OpenAI's ChatGPT 3.5 to ChatGPT 4 helped to drive the AI boom, as the technology went from an impressive chatbot to a technology seemingly capable of listening, seeing, and understanding the world around it. When OpenAI showed off its 4o "omni" upgrade this time last year, the company described it as "a step towards much more natural human-computer interaction."

Fast forward to today, and the Wall Street Journal is reporting that Meta has delayed the release of its Llama 4 "Behemoth" AI model because its gains over previous iterations were not as impressive as the company had hoped.

"The Facebook-parent has publicly touted the capabilities of Behemoth, saying it already outperforms similar technology from OpenAI, Google and Anthropic on some tests," the WSJ reported. "But internally, its performance has been hobbled by training challenges."

OpenAI has faced similar challenges with its work-in-progress next upgrade, the WSJ previously reported.

It's hard to tell what all these delays for next-gen AI will mean, though. Even at its current capabilities, current AI is already changing the way we interact with computers at work and at home.

Data increasingly shows AI is upending the worlds of medicine, art, coding and even Google searches. The changes are so pronounced that last week, an Apple executive testified in court that Google searches through the built-in iPhone, iPad and Mac web browsers had dropped for the first time -- ever -- last month. So, now Apple is "actively looking at" changing its web browser to focus on AI.

While Facebook is working on its slow-growing Llamas...

OpenAI releases GPT-4.1 to all ChatGPT users

OpenAI announced on social media that its latest upgrade, GPT-4.1, would soon be available to everyone using ChatGPT. The technology, which was previously only available to developers, is reportedly optimized in ways that help it excel at tasks like coding. OpenAI said GPT-4.1 can also understand more text, video, audio or images fed to it through a prompt than previous iterations.

Alongside those moves, OpenAI also released safety testing information about GPT-4.1 publicly on a new "Safety evaluations hub," providing details about testing around harmful content, hallucinations and other issues.

That's all well and good, but it may be too late, because the ultimate AI has arrived...

Researchers create LegoGPT

Yup, that's right, Carnegie Mellon University has cracked the nut. It has done what was once unthinkable. It has changed how we think of computers forever. It has created a Lego AI.

Jokes aside, this is much more than the image generating trick where you just ask ChatGPT to reimagine you as something that looks like a Lego creation. Often, those images create other-worldly Lego pieces that you can't find in the mortal realm. But LegoGPT is designed to create images of pieces that you can actually build.

"Our experiments show that LegoGPT produces stable, diverse, and aesthetically pleasing LEGO designs," the researchers wrote. "Our designs can be assembled manually by humans and automatically by robotic arms."

So, that's it. Show's over. Humanity has reached its apex. You can all go home now.

On your way, you might want to consider...

Waymo recalls 1,200 of its robotaxis

This might be a forehead slap moment for many of us. The company, which used to be known as Google's self-driving car project from the Google X research lab, has reportedly recalled 1,200 of its driverless taxis after reported crashes with some road barriers, like gates.

A Waymo spokesperson reportedly said the recalls "are not affecting any vehicles currently on the road," according to CNET. Waymo robotaxis currently run in San Francisco, Los Angeles, Phoenix and Austin.

More Gemini more places

Ahead of the Google I/O developer conference next week, the company announced its next major Android update, which of course includes a bunch more Gemini AI everywhere. Soon, the company said, Gemini will be available on wristworn Wear OS devices, as well as the company's new Android XR headsets, TV interface and even Android Auto.

"We’ve completely reimagined Android with Gemini right at the center," Google said in a statement on its site. "We’re expanding Gemini way beyond phones, bringing a truly smart and helpful assistant to all your devices."