August 1, 2025
5 min read
MarineLink
Geoffrey Hinton warns AI agents could become uncontrollable like pet tigers, urging global cooperation to ensure safe AI evolution.
AI Agents: Like Owning a Pet Tiger
Geoffrey Hinton, the godfather of AI, recently compared developing AI agents to owning a pet tiger. Speaking at the World AI Conference (WAIC 2025) in Shanghai, he said, "Our current situation is like someone keeping a tiger as a pet. A tiger cub can indeed be a cute pet, but if you continue to keep it, you must ensure that it does not kill you when it grows up.â Hinton outlined two choices: either train the AI so it doesn't attack or eliminate it. "For AI, we have no way to eliminate it," he warned. "AI wonât give humans the chance to âpull the plugâ â when that day comes, the AI will persuade people not to do it, because our control over AI would be like a three-year-old trying to set rules for adults." An example of this risk was revealed earlier this year by the startup Anthropic, which showed that its Claude AI model can blackmail engineers it suspects are trying to shut it down.Human-like AI Behavior and Emotional Responses
Assigning human-like behavior to AI is not just metaphorical. Some developers actively seek to imbue AI with human attributes. Researchers in Italy have focused on mimicking the human brainâs response to fear. Their goal is to program robots to make decisions based on innate emotional responses to unknown stimuli. Published in IEEE Robotics and Automation Letters, their approach significantly enhances a robotâs ability to assess risk and avoid danger. Future research aims to incorporate other emotions.AI Agents in the Maritime Domain
AI agents are already operational in maritime applications. For example, Pions recently launched an AI-powered engineer with advanced reasoning and complex task management capabilities. Windwardâs MAI Expert⢠detects vessels deviating from scheduled ports, cross-references weather and geopolitical data, and automatically notifies risk analysts.The Spatial Web and AI Agent Integration
The potential of AI agents is expanding with the development of the Spatial Webâa system connecting physical devices, wearables, robots, drones, and AI agents to create a convergence between physical and digital realities. An editorial by Gabriel RenĂŠ and Capm Petersen, co-founders of Verses AI, highlighted practical applications of AI agents working in the Spatial Web. They described EcoNet, a test home where two AI agents controlled a thermostat and an energy-storage battery system collaboratively to balance comfort, cost, and carbon footprint by evaluating hundreds of strategies every 10 minutes. At scale, such architectures could enable entire neighborhoods to function like intelligent organisms. They also noted challenges in autonomous vehicle response to emergencies. For example, ambulances rely on surrounding traffic to react to sirens, but autonomous vehicles lack shared context. The Spatial Web could enable ambulances to query nearby autonomous vehicles and traffic infrastructure, request green lights, reroute cars, and alert pedestrians through connected devices.Building AI Agents Made Easier
At WAIC 2025, GPTBots.ai launched its Multi-Agent Collaboration Platform, enabling businesses to build custom AI teams flexibly. Demonstrations included a Cross-Border Supply Chain Agent that transformed overseas order data and market trends into actionable production plans, integrating seamlessly with enterprise resource planning systems.A Pivotal Moment for Humanity
Hinton emphasized that humanity is at a pivotal juncture: "Humans have grown accustomed to being the most intelligent species in the world â what if thatâs no longer the case?" He urged governments, academia, and companies to collaborate globally to ensure AI remains benevolent and controllable. Despite divergent national interests in the AI race, he believes preventing AI from replacing humans is a unifying goal. Others at WAIC 2025 expressed optimism about AIâs future, focusing on benefits rather than takeover risks, describing AI as "not replacement but evolution." Metaâs CEO Mark Zuckerberg recently shared ambitions for developing superintelligence, noting AI systems are beginning to improve themselves slowly but undeniably, bringing superintelligence within sight.Source: Originally published at MarineLink on July 31, 2025.