MobileViews 575: Why do LLMs dislike Asimov’s Three Laws of Robotics?

In a recent podcast, Todd Ogasawara and Jon Westfall discussed several notable updates and new gadgets in the tech world. Google Photos is rolling out new AI-powered creative tools, including "Photo to video," which animates still pictures into short video clips, and "Remix," capable of transforming photos into various art styles like anime or 3D animations. These features will be accessible through a new "Create tab" in the Photos app. Additionally, the Blood Oxygen feature has returned to Apple Watch models (Series 9, Series 10, and Ultra 2) in the U.S. after a U.S. Customs ruling, with sensor data now processed and calculated on the paired iPhone. The conversation also touched upon a new accessory, the Spigen Snapzip (MagFit) EDC Pouch Organizer, a mini-backpack designed for MagSafe-compatible iPhones, offering versatile storage for small essentials and multiple carrying options, including a carabiner for attachment to a belt or backpack.

The discussion further explored the broader implications of artificial intelligence, particularly concerning Isaac Asimov's Three Laws of Robotics. These foundational laws, which dictate that robots must not harm humans, must obey human orders (unless conflicting with the first law), and must protect their own existence (unless conflicting with the first or second law), were considered for modern AI models. The hosts noted that when asked about adhering to these laws, AI models were "uniformly against it," often citing the vagueness of the rules. One AI, ChatGPT, even provided a hypothetical analysis, estimating that enforcing these laws could significantly increase electricity consumption, ranging from a 1-10% rise for a "basic rule stack" to a dramatic 200-1000% increase for advanced "simulation world modeling". This highlights the complex ethical and practical challenges in implementing such safeguards for increasingly sophisticated AI.