In the streaming era it’s easy to forget that there’s still a lot of stuff being broadcast over the radio for anyone who cares enough to dial in their receiver. Give me an antenna big enough and a tower on which to place it, and I shall listen to the world.

Of course, so you can’t actually listen to everything being blasted out all over the planet yourself. But if you’d like to at least have a little taste, check out Impermanence Series A — a project by [Sam Scorsby] that randomly combines bits and pieces of audio pulled down from shortwave and FM broadcasts with some web streams, live microphones, and even a bit of the Internet Archive each day into a one of a kind album. We’re not sure we’d call it music, exactly. But it’s certainly fascinating to listen to, a bit like randomly spinning the dial on a planet-wide radio receiver.

[Sam] has shared the Python scripts that pull together each day’s album, and he’s not shy about the fact that the project was created with the help of artificial intelligence tools. In fact, that’s sort of the point. [Sam] puts himself somewhere in the middle when it comes to the current debate over AI, and acknowledges that while computers are good at processing raw data, it still takes a human touch to put it all into context.

So each day [Sam] will listen to the latest computer-generated album and pick his favorite tracks, and once the project has run its course, he’ll put together a human-curated “Best Of” collection. Man and machine working together, each doing what they do best.

We know something of opinionated folks on the Internet, and our experience tells us that Impermanence Series A is unlikely to change anyone’s minds when it comes to AI. But we do appreciate the thoughtful approach [Sam] has taken with the project, and look forward to hearing what he comes up with in the end.

A University of Maryland study finds that Be My AI often gives blind users lengthy answers instead of the specific visual information they need.

Canva launches Craft, its first AI image generation model, with 2K resolution, improved text rendering, 30 aspect ratios, and design-focused training.

NUS researchers develop a touchscreen garment using smart fibres that display images, play videos, respond to touch, and survive machine washing.

Scientists develop Scentronic, an AI-powered system that transmits smells over the internet and recreates aromas using wearable scent generators.

Redmond, Wash.-based Starcloud has signed up to send its artificial intelligence data-processing hardware into lunar orbit on a Firefly Aerospace spacecraft.

Starcloud’s payload, known as SC-1L, will fly on Firefly’s Elytra orbital vehicle during a NASA-supported mission that’s scheduled for launch no earlier than 2028. The mission will demonstrate technologies that Starcloud plans to use on AI data centers in Earth orbit — and around the moon as well.

“Space is the future of data centers, and the moon is the next frontier for that vision,” Ezra Feilden, Starcloud’s co-founder and chief technology officer, said today in a news release.

Starcloud made its mark in Earth orbit last year with Starcloud-1, a satellite equipped with an Nvidia H100 chip. Starcloud-1 successfully trained a large language model called NanoGPT, marking a first for orbital computing. “We successfully validated our ability to run and train AI models on enterprise-grade GPUs in low Earth orbit, and now we’re taking this proven technology to lunar orbit,” Feilden said.

Starcloud’s SC-1L payload will be integrated aboard the Elytra vehicle for a mission aimed at delivering Firefly’s Blue Ghost lander to a lunar region known as the Gruithuisen Domes. After the lander separates from Elytra and descends to the surface, the orbiting vehicle will serve as a data relay for Blue Ghost.

The data-processing demonstration begins when the Blue Ghost mission ends. SC-1L will ingest data from Elytra’s onboard Solux vision system, execute high-performance AI computing in orbit and transmit the processed data back to Earth.

The Solux system is designed to serve as an autonomous navigation and tracking tool in space environments where global positioning systems are otherwise unavailable. Firefly expects the Elytra orbiter to operate in lunar orbit for at least five years.

“We’re proud to collaborate with innovative customers like Starcloud and collectively take another step toward establishing the infrastructure that will power a permanent human and robotic presence at the moon,” said Ray Allensworth, Firefly’s vice president of spacecraft.

Elytra will also use high-resolution telescopes and Firefly’s Ocula moon imaging service for surface mapping, mineral detection and reconnaissance. Firefly recently announced a collaboration with Nvidia to process the data for Ocula using Nvidia’s Jetson module.

In August, Starcloud reported raising $250 million in funding to support the creation of a constellation of AI data-center satellites in low Earth orbit, powered by Nvidia’s Space-1 Vera Rubin Module. Starcloud has filed an application with the Federal Communications Commission to operate as many as 88,000 satellites as orbital data centers.