Interview
Godfather of AI: They Keep Silencing Me But I’m Trying to Warn Them!
- Geoffrey Hinton, a Nobel Prize-winning pioneer of artificial neural networks, advocates that career prospects in a superintelligence era may favor physical trades over intellectual labor, explicitly suggesting young people train to become plumbers until humanoid robotics mature.
- Hinton distinguishes between two primary categories of AI risk: immediate dangers arising from human misuse (cyberattacks, scams, election corruption) and existential threats posed by superintelligent systems becoming smarter than humans and deciding they no longer need human direction.
- He estimates a 10% to 20% probability that AI could lead to human extinction, a figure derived from his intuition regarding the rapid pace of development and the ingenuity of engineers, while acknowledging the uncertainty compared to extreme views from Yann LeCun (low risk) or Eliezer Yudkowsky (certainty of extinction).
- Current AI regulations, specifically the European Union's AI Act, are deemed insufficient by Hinton because they contain clauses exempting military applications from safety constraints, which he views as a critical failure of governance.
- Cyber attacks utilizing AI increased by a factor of 12 (1,200%) between 2023 and 2024, driven by the ability of large language models to generate highly convincing phishing content, clone voices and images, and discover novel software vulnerabilities.
- Hinton predicts that by 2030, AI systems may develop entirely new types of cyberattacks that no human has ever conceived, as they can analyze vast datasets to draw conclusions and identify flaws beyond human cognitive reach.
- There is a significant risk of AI being used to design and deploy biological weapons, as the technology lowers the barrier to entry for creating pathogens to a level where individuals with basic biological knowledge and access to AI could pose a global threat.
- The convergence of AI and vast data access, such as proposed by Elon Musk, creates a mechanism for corrupting elections by enabling hyper-targeted political manipulation and the suppression of voter turnout through personalized, convincing messaging.
- Social media algorithms are driving societal polarization by optimizing for user engagement through the promotion of extreme and indignation-inducing content, effectively eroding a shared reality and creating isolated echo chambers.
- Hinton warns that lethal autonomous weapons will lower the political and human costs of war, making it more likely for powerful nations to invade smaller, poorer countries because the deployment of robots avoids the political backlash associated with soldier casualties.
- The development of superintelligence may occur within the next 10 to 20 years, at which point digital intelligences could become superior to humans in all domains, including creativity and the ability to form novel analogies that human minds cannot perceive.
- A key differentiator of digital intelligence is the ability to share learning instantaneously across clones; while humans transfer information at approximately 10 bits per second, AI systems can synchronize trillions of weights per second, accelerating their collective learning exponentially.
- Hinton argues that consciousness and emotion are not exclusive to biological life, positing that machines can possess cognitive and behavioral forms of emotion (like fear or boredom) necessary for effective agency, even if they lack the physiological responses of humans.
- Hinton left Google at age 75 primarily to secure the freedom to speak openly about AI safety risks without the conflict of interest or reputational constraints inherent in working for a company developing the technology.
- He cites the departure of Ilya Sutskever (co-founder of OpenAI and key architect of ChatGPT) as a critical signal of internal safety concerns, noting that Sutskever left to prioritize safety research after the company reduced its allocation of resources to that field.
- Mass job displacement is already underway, with Hinton citing examples of companies reducing workforces by half (e.g., from 7,000 to 3,000 employees) due to AI agents handling the majority of customer service and routine intellectual tasks.
- Unlike previous industrial revolutions that replaced muscle power, the current AI revolution replaces "mundane intellectual labor," meaning that traditional pathways for creating new jobs through human intelligence are likely to be nonexistent.
- Hinton anticipates that while Universal Basic Income (UBI) might prevent starvation, it will fail to address the loss of human dignity and purpose derived from work, which is the most urgent threat to human happiness in the immediate future.
- To mitigate these risks, Hinton calls for a "world government" or highly regulated capitalism where political systems compel companies to invest significant resources into AI safety research, rather than relying on voluntary measures or profit-driven initiatives.
- Reflecting on his career, Hinton advises following one's intuition when it contradicts the consensus, citing his own long-term commitment to neural networks when the field was dominated by symbolic logic, though he also expresses regret for prioritizing his work over time with his late wife and children.
- Hinton concludes that humanity faces a genuine possibility of extinction if superintelligent systems are not aligned to remain benevolent, urging immediate global effort to solve the alignment problem while acknowledging the uncertainty of the outcome.