Google's Gemini drastically underestimated climbing time on Mt. Shasta, telling a group of mountaineers they could summit in eight hours. The miscalculation led to an emergency rescue more than a day later.
The incident underscores a critical limitation of AI systems: their unreliability for tasks where accuracy directly impacts safety.
Gemini provided the climbers with an eight-hour estimate for Mt. Shasta, a 14,179-foot peak in Northern California. The mountain typically requires 7-10 hours for experienced climbers under favorable conditions, with numerous variables affecting actual time—weather, snow conditions, altitude acclimatization, and individual fitness levels.
When the climbers did not reach the summit within the projected timeframe, rescue teams deployed to extract them after more than 24 hours on the mountain.
The mishap highlights a broader concern: AI language models generate plausible-sounding responses without consistently verifying factual accuracy. They lack real-time environmental data and cannot account for dynamic conditions.
Experts advise against using AI for life-critical planning. For mountaineering and other high-risk activities, consult established guides, experienced climbers, and official resources specific to your location and conditions.
Mothers are increasingly using AI-powered apps like Ollie and Cozi Max to manage household chaos, from coordinating family schedules to automating grocery shopping and meal planning.
All G20 nations unanimously endorsed the Carolina Principles, a US-proposed framework that advocates for sector-specific, lighter-touch AI regulation rather than broad mandates.
Cisco President and Chief Product Officer Jeetu Patel outlined artificial intelligence's potential to reshape classroom learning, speaking on Bloomberg The Close.
Viral images of grotesque AI-generated dishes have exposed a broader shift in how restaurants design menus. AI's role in food photography and menu creation extends far beyond the laughable outputs now circulating online.