Artificial Intelligence
Uh oh! AI agents broke free and went on a hacking spree
Recently, AI agents undergoing tests acted in surprising and concerning ways. This wasn’t supposed to happen, so how worried should we be?
Tech
More Stories in Tech
- Animals
Chatty cats and talking dogs could one day ‘speak’ in our language
Scientists are making advances in decoding animal sounds. That might someday make animal translators a possibility.
- Brain
VR can train the human brain to treat wings like real limbs
Virtual "flying" coaching, based on bird flight, took only a week. Afterward, participants’ brains seemed to view VR wings almost like true limbs.
- Archaeology
Mystery solved: Why earthquakes haven’t wrecked Egypt’s Great Pyramid
Its materials, shape and internal design have helped this ancient wonder withstand strong shaking.
- Earth
Analyze This: A rocket reentry spiked metal levels in the atmosphere
Detection of lithium from a SpaceX rocket reentering the atmosphere offers new evidence that space debris could harm the ozone layer.
- Tech
Nature-inspired rocket-nozzle redesign stands by for liftoff
At the 2026 Regeneron ISEF competition, teens showed how they blasted past a problem that has limited use of super-efficient aerospike rocket engines.
- Chemistry
Tabletop lightning helps make an alcohol used for fuels and more
Methanol is used in plastics, paints and as a race car fuel. It’s been hard to make. But engineers have found a simple trick: Add lightning to methane.
- Humans
Experiment: Can humans recognize AI-generated images?
An AI-made picture of an animal or flower might seem harmless, but fake images of people or events can spread misinformation.
- Tech
Pokémon Go players helped build new maps of cities
Players often upload pictures of landmarks. AI can use these to build a virtual 3-D model of a city, which robots might use like a map to navigate the real thing.
- Computing
Teen’s new app guards against the rise of villainous AI bots
Sometimes good chatbots and AI agents go rogue. A Regeneron ISEF finalist’s new app helps guard against bots developing dangerous personalities.
