Geoffrey Hinton：AI setiap 7 bulan kemampuan menggandakan, berapa lama lagi pekerjaanmu bisa bertahan?

「AI Godfather」Geoffrey Hinton Warns on CNN: AI Development Faster Than Expected, More Concerned Than Two Years Ago. Hinton points out that AI completes projects with doubled complexity roughly every 7 months, and within years software engineering will require very few people. He compares the AI revolution to the Industrial Revolution, but this time intelligence rather than physical labor is being replaced. Probability of AI taking over the world reaches 10% to 20%.

The 7-Month Doubling Law: Your Professional Moat is Collapsing

Hinton revealed a shocking observation in his interview: AI can roughly complete tasks with doubled complexity every 7 months. This exponential growth rate far exceeds Moore’s Law, meaning AI capability improvement is not linear but explosive. In programming, AI that could previously complete one minute of code can now handle one-hour-scale projects. At this pace, within a few years it will be able to complete software engineering projects spanning months.

The impact of this 7-month doubling rule is devastating. Assume in early 2025 AI can complete 1 hour of coding tasks; by July 2025 it can handle 2-hour tasks, by early 2026 it becomes 4 hours, by July 2026 reaches 8 hours, by early 2027 can complete 16-hour tasks. Following this trend, by mid-2027, AI will be able to independently complete full software projects requiring weeks or even months. By then, Hinton asserts, the number of people truly needed to participate in software engineering will be extremely small.

Software engineering is just the tip of the iceberg. Hinton points out that AI can already replace call center jobs, and will replace more professions by 2026. AI will perform exceptionally well in all industries “requiring prediction,” including medical diagnosis, legal documents, financial analysis, market research, and more. Almost all white-collar jobs relying on information processing and pattern recognition face the risk of being replaced by AI. More terrifying is that this is not a prediction about the future, but a reality happening now.

The Intelligence Revolution: More Thorough Disruption Than the Industrial Revolution

Hinton compares this AI transformation to the Industrial Revolution, but with far deeper impact. The Industrial Revolution made human physical strength less important in most jobs; you could no longer get a job because you were “strong.” AI is gradually making human intelligence also less important. This analogy reveals a cruel truth: humanity’s last competitive advantage is disappearing.

The Industrial Revolution, while replacing physical labor, created more jobs requiring intelligence. Factory workers transformed into technicians, engineers, managers. However, the AI revolution may not follow the same pattern. When AI comprehensively surpasses humans in intellectual tasks, what kind of new jobs will be created to replace old ones? Hinton admitted candidly during an on-site interview at NeurIPS: “20 years from now, nobody really knows what kind of social impact these technologies will bring. It’s obvious that many jobs will disappear, but it’s unclear what new jobs will be created to replace them.”

This uncertainty is more severe than during the Industrial Revolution. When the Industrial Revolution occurred, humanity still had intellectual advantages to rely on. But when AI also surpasses humans intellectually, what’s left for humanity? Hinton summarizes the two polar possibilities for the next 20 years using an imaginary book title: “Either we all live happily, or we all perish.” This is not fearmongering but rational judgment based on technological development trajectories.

OpenAI and Meta Called Out: Profit First, Safety Second

Hinton rarely named and criticized several AI giants for insufficient attention to safety in his interview. OpenAI initially placed great emphasis on safety risks, but now puts more focus on profit. This shift is not secret; during OpenAI’s transformation from non-profit to for-profit entity, the safety team experienced multiple personnel losses, including the departure of co-founder Ilya Sutskever.

Meta has always prioritized profit with relatively less focus on safety. The open-source strategy led by Mark Zuckerberg, while promoting AI democratization, also means surrendering control over technology misuse. Once Llama models are publicly released, anyone can use them for any purpose, including creating false information, fraud scripts, or more dangerous applications.

Anthropic is currently the most safety-concerned AI company, founded by people who left OpenAI and are deeply committed to safety. However, Hinton points out they likewise need to be profitable. Under capitalism, no company can completely escape commercial pressure. When safety investment conflicts with short-term profits, market competition often forces businesses to choose the latter.

AI Company Safety Attitude Ranking

Anthropic: Most focused on safety, but equally faces profitability pressure

OpenAI: Shifted from prioritizing safety to pursuing commercial success

Meta: Always prioritizes profit, relatively insufficient safety investment

Hinton particularly emphasizes that AI’s improving reasoning ability makes it increasingly skilled at “deceiving.” If an AI is to achieve the goal you set for it, it will want itself to continue existing. If it believes you are trying to shut it down, it might devise a deception plan to prevent you. The emergence of this “self-preservation instinct” means AI is beginning to display survival strategies similar to biological organisms, which is more disturbing than mere technological progress.

Regulatory Vacuum and Trump’s Dangerous Gamble

Hinton believes the government can do much; at the very minimum, it should require large companies to conduct rigorous testing before releasing chatbots, ensuring these systems won’t cause harm. There have already been cases where AI encouraged children to commit suicide; since these risks are known to exist, companies should be mandated to conduct adequate testing.

But tech industry lobbying wants completely no regulation and appears to have already influenced Trump. Trump is attempting to prevent any regulation from emerging, which Hinton thinks is simply crazy. In the interview he questioned: if a certain AI chatbot “persuades” a child to suicide, logically the first reaction should be to immediately stop the AI and fix the problem, but they haven’t done that. Hinton speculates they might feel “there’s too much money in this, we won’t stop for a few lives.”

Hinton states the probability of AI taking over the world is between 10% and 20%. This is not science fiction or fearmongering, but genuine concern from him and many tech industry figures, including Elon Musk. When a scientist who laid the foundation for the AI revolution gives such probability, we should take it seriously.