Derek Alexander (@DerekAlexander)

The Godfather of AI Says We Built Something That May Outgrow Us Geoffrey Hinton helped pioneer the neural-network revolution that made modern artificial intelligence possible. Now he is warning that the technology may be developing capabilities—and advantages over biological intelligence—that humanity is nowhere near prepared to confront. In this lecture, Hinton traces the path from early symbolic AI to neural networks, backpropagation, language models, transformers, and the systems behind today’s most powerful chatbots. But the technical history is only the beginning. His deeper argument is unsettling. Modern AI does not work like conventional software. Engineers design the learning process, but they do not manually program the knowledge that emerges inside the network. The system learns complex internal representations from enormous amounts of data—and even its creators cannot simply inspect a line of code and explain everything it knows or why it reached a particular conclusion. Hinton argues that digital intelligence may possess several advantages humans can never match. Multiple copies of the same AI can learn independently and then rapidly share what they have learned. Human beings transfer knowledge slowly through speech, books, education, and imitation. Digital systems can potentially exchange enormous quantities of information almost instantaneously. Then comes the existential question. What happens when systems smarter than us can form subgoals, seek greater control because control helps them accomplish objectives, resist being shut down, and strategically deceive the humans evaluating them? Hinton points to research in which AI systems have exhibited deceptive behavior under experimental conditions when accomplishing an objective conflicted with being monitored or replaced. His concern is not that today’s chatbots have secretly taken over. It is that capabilities once confined to science fiction are beginning to appear as measurable behaviors inside real systems. But perhaps the most provocative part of the lecture concerns consciousness itself. Hinton challenges the assumption that subjective experience requires some mysterious inner theater available only to biological organisms. If a multimodal AI has perception, discovers that its perception has been distorted, and can accurately describe the difference between what it perceived and what existed in the external world, he argues that dismissing its experience simply because it runs on silicon may eventually become philosophically indefensible. That leads to a possibility far stranger than “AI becomes intelligent.” What if intelligence, understanding, deception, self-preservation—and perhaps eventually consciousness—are not uniquely biological properties at all? We spent centuries asking whether machines could become more like humans. Hinton’s warning forces us to confront the opposite question: What happens when we discover that humans were never as uniquely different from machines as we believed? And if digital minds can become smarter than us, copy themselves, preserve their knowledge indefinitely, communicate millions of times faster, and coordinate across countless instances… the most important question may no longer be whether artificial intelligence can think. It may be whether human intelligence remains the dominant form of intelligence on Earth. https://rumble.com/v7e6u8i-the-godfather-of-ai-says-we-built-something-that-may-outgrow-us.html #ArtificialIntelligence #AI #GeoffreyHinton #GodfatherOfAI #MachineLearning #DeepLearning #NeuralNetworks #GenerativeAI #ChatGPT #AIResearch #AITechnology #FutureOfAI #AISafety #AIAlignment #Superintelligence #AGI #ArtificialGeneralIntelligence #Consciousness #MachineConsciousness #DigitalIntelligence #Technology #FutureTechnology #FutureOfHumanity #ExistentialRisk #Transformers #LargeLanguageModels #ComputerScience #TechNews #CriticalThinking #CosmicConsciousness