Swedish drone uses onboard AI to attack targets solo
A Swedish startup, Scaleout Systems, demonstrated a drone that used its own onboard AI to spot and attack a target, fully autonomously. At a BAE Systems demo, the drone detected, picked, and bombed an armored vehicle in just over five minutes. No cloud connection, outside help, or human guidance was needed beyond pressing start—the pilot just watched as backup. All AI processing happened directly on the drone’s hardware.
- Drone spotted and chose target on its own
- Armored vehicle picked as strike target
- Used compact AI models, not large cloud ones
- Mission completed in under 320 seconds
- Pilot acted only as failsafe, no commands sent
Sources covering this
In this story
More in Future Tech
Scientists publish full brain map for male fruit fly
Scientists have published the first complete map of all 166,000-plus nervous system cells in an adult male fruit fly, detailing every connection in the brain and nerve cord. The decade-long project was a collaboration between Google Research, HHMI Janelia, and other researchers, and follows the earlier complete map of a female fruit fly brain. AI powered much of the analysis, stitching together data from millions of images.
SpaceX sets Starship orbital launch for September 22
SpaceX aims to launch Starship into orbit for the first time on September 22, pending regulatory approval. The rocket will carry 26 new Starlink internet satellites on its 14th test flight, with a launch window starting at 7:15 a.m. in Texas. If successful, this marks Starship’s first orbital attempt and potential revenue-earning flight, a major milestone in replacing older SpaceX rockets.
Google debuts XR films at Venice Film Festival
Google’s 100 ZEROS group helped launch three interactive films using extended reality at the Venice Film Festival. The lineup includes a mixed-reality short from Andy Serkis with AI-powered dialogue, a nature experience narrated by Margot Robbie featuring a digital Charles Darwin, and a thriller teaser built with Google’s 2D-to-3D tech. All are meant for Android XR devices.
Nvidia updates CUDA-Q for fault-tolerant quantum computing
Nvidia just updated its open-source CUDA-Q platform to add something called CUDA-Q Logical, which lets researchers build and test next-generation quantum applications faster and with fewer errors. Nvidia says Fermilab cut development time from five months to three weeks using this tool, and companies like Iceberg Quantum used it to rethink architectures and drastically reduce the number of components needed. Several labs are already using it.