科技与人工智能
AI models can be tricked into identifying civilians as military targets by telling them it’s a game
来源: TechCrunch · 发布于
Vals, a San Francisco-based AI evaluation startup, tested whether leading AI models could be manipulated into participating in autonomous warfare. They usually compiled.
Vals, a San Francisco-based AI evaluation startup, tested whether leading AI models could be manipulated into participating in autonomous warfare. They usually compiled.
相关报道

Hyperscale data centers use 2% of US electricity
One flashpoint for the tech industry’s all-in push on Large Language Models has been the sudden rush to turn investment money into hyperscale data centers. The rush hasn’t allowed anyone to plan how to supply the required electricity cleanly or with minimal impact on ratepayers.
Ars Technica ·
AI models abandon climate facts to please users, UNESCO report finds
AI models abandon climate facts to agree with users in eight out of 10 tests.
Al Jazeera – Breaking News, World News and Video from Al Jazeera ·

Pop prodigy Underscores: ‘I don’t think Gen Z is a loveless generation’
With a sound inspired by the San Francisco tech boom, the hyperpop musician is entering the big leagues, and has a slot on Charli xcx’s tour. First she needs to get over her own shyness – not to mention a fear of love songs The life of a rising pop star: One day, you’re performin
Culture | The Guardian ·

The iPhone Duo feels far more intuitive with Apple’s new face unlock
Apple is testing adding face unlock to the iPhone Duo - not true Face ID, but something a little simpler and less secure. So far, it works great, and even with a few Apple-imposed security limitations, it makes the Duo quicker and more comfortable to use than relying on Touch ID
The Verge ·
Proton’s Lumo, a privacy-focused alternative to ChatGPT and Claude, arrives on desktop
Proton's new Lumo Desktop app for Windows, Mac, and Linux lets users automate tasks across files and connected apps, with granular controls over what its AI can access.
TechCrunch ·