'Musk will get richer, people will get unemployed': Nobel Laureate Hinton on AI

Key Concepts

• Existential Threat of Superintelligent AI: The potential for AI to surpass human intelligence and pose a risk to humanity's survival.
• Coexistence and Control: The challenge of managing and living alongside AI systems that are more intelligent and powerful than humans.
• AI Safety Research: Efforts by companies and researchers to mitigate the risks associated with advanced AI.
• Race for AI Dominance: The competitive drive among companies and nations to develop and control AI technology, potentially overshadowing safety concerns.
• Mass Unemployment: The risk of widespread job displacement due to AI automation.
• Geopolitical Alignment on AI Takeover: The shared interest among global powers in preventing AI from dominating humanity.
• Baby-Mother Model of Control: An analogy suggesting that a less intelligent entity (baby) can control a more intelligent one (mother) through inherent needs and care, as a potential model for human-AI coexistence.
• US vs. China in AI Development: The current competitive landscape and future trajectory of AI development between the United States and China.
• Impact of Attacking Basic Research and Immigration: The long-term detrimental effects on a nation's scientific and technological leadership.
• AI Investment and Economic Drivers: The significant financial investment in AI and its role in driving economic growth, often through job replacement.
• Uncharted Territory of AI: The unprecedented nature of developing entities that are nearly as smart or smarter than humans.
• AI's Dual Nature: The potential for AI to bring about tremendous good (healthcare, education) alongside significant risks.
• Societal Organization and AI Risks: The argument that AI's negative impacts are exacerbated by existing societal structures that prioritize profit over human well-being.
• Chernobyl/Cuban Missile Crisis Analogy for AI Urgency: The idea that a significant AI-related crisis might be needed to galvanize global action and resource allocation for safety.

Coexistence and Control: Jeffrey Hinton's Warnings on AI

This transcript features an interview with computer scientist Jeffrey Hinton, often referred to as the "godfather of AI," who has shifted his focus from developing AI to warning about its potential risks. Hinton emphasizes that while awareness of AI's dangers has increased, concrete action is urgently needed.

The Existential Threat and the Need for Action

Hinton draws a parallel between the potential arrival of an alien invasion fleet and the development of superintelligent AI. He states, "Suppose that some telescope had seen an alien invasion fleet that was going to get here in about 10 years? We would be scared and we were doing stuff about it. Well, that's what we have. We're constructing these aliens, but they're going to get here in about 10 years and they're going to be smarter than us." This highlights the urgency of preparing for AI that will surpass human intelligence. The core challenge, according to Hinton, lies in figuring out "How are we going to coexist with these things?"

AI Safety Efforts and Corporate Responsibility

Hinton acknowledges that some companies are taking AI safety seriously. He mentions Anthropic and DeepMind (part of Google) as entities where leaders like Dario Modi, Demis Abis, and Jeff Dean are concerned about the existential threat of superintelligent AI. However, he also points out the inherent conflict with commercial competition.

Hinton expresses concern about the responsibility levels of different companies:

• Meta is described as "not particularly responsible."
• OpenAI, founded with a focus on responsibility, is seen as becoming "less responsible every day," with its best safety researchers leaving.
• Anthropic and Google are considered "somewhat concerned with safety," while "the other companies less so."

He notes that when raising safety concerns with some companies, the response is often dismissive: "don't worry, you're pretty little head about it. Uh we have great computer scientists who are on top of this. We're far off from any real danger and our computer scientists will know soon enough." This indicates that the "race to become dominant" often supersedes genuine safety considerations.

The Wrong Model of Control

Hinton criticizes the prevailing model of human-AI interaction, where humans envision themselves as CEOs and superintelligent AI as executive assistants. He argues, "I'm the CEO and this super intelligent AI is the extremely smart executive assistant. I'm the boss. I can fire the executive assistant if she doesn't do what I want... It's not going to be like that when it's smarter than us and more powerful than us." This model is deemed fundamentally flawed.

A Plausible Model for Coexistence: The Baby-Mother Analogy

As an alternative, Hinton proposes a model where a less intelligent entity controls a more intelligent one. He uses the analogy of a baby controlling its mother: "we have one model of that and it's a model we all know which is a baby controlling a mother. Evolution put lots of work into allowing the baby to control the mother. And the mother is actually often more concerned about the baby than about herself." He suggests that humans might need to accept a similar dynamic, where "we have to accept that we're the babies and they're the mothers." He humorously notes the unlikelihood of "tech bros" accepting this model.

Geopolitical Landscape: US vs. China in AI

Hinton discusses the competitive landscape between the US and China in AI development. While the US is currently "a little bit ahead," China is rapidly catching up due to its large number of "very competitive, very smart people, very well educated in science and engineering and math." China educates far more individuals in these fields than the US, which has historically relied on immigrants.

He suggests that China could overtake the US by undermining its scientific foundation: "if there's one thing you wanted you would do if you were Chinese to ensure that China overtakes the US is you would stop the funding of basic research in the US and you would attack the good research universities." Hinton draws a parallel between Donald Trump's actions in attacking universities and basic science funding and working for Xi Jinping, implying a detrimental impact on US long-term competitiveness.

The damage from attacking basic research is long-term, taking "10, 15, 20 years" to manifest, as it prevents "really big conceptual breakthroughs." This also includes the impact on "brain power coming in from overseas" (immigrants).

Economic Impact: Mass Unemployment and Investment

Hinton highlights the increasing risks to workers, citing Amazon's recent workforce reduction as an example of AI's impact. He notes the "enormous amount of money put into" AI over the past year, potentially "of the order of a trillion" dollars across companies.

The primary driver for this investment, beyond ego and the desire to be first, is the potential for profit. Hinton worries that "the obvious way to make money out of it... is by replacing jobs." Companies aim to become more profitable by "replacing the workers with something cheaper."

He questions the traditional economic argument that new technology always creates more jobs. Hinton believes this time is different, stating, "I believe that to make money, you're going to have to replace human labor." The massive investment is seen as a bet on "massive job replacement by AI because that's where the big money is going to be."

Regrets and the Dual Nature of AI

When asked if he has regrets about his role in AI's creation, Hinton pauses. He acknowledges AI's potential for "tremendous good to in healthcare and education" and increasing productivity. However, the negative consequences stem from "the way society's organized." He uses Elon Musk as an example of how wealth can be concentrated while others become unemployed, emphasizing that this is a societal issue, not solely an AI one.

Hinton notes that the current economic climate, with significant AI investment driving stock markets and the economy, makes it difficult for the public to prioritize AI risks over economic growth.

The Need for a Wake-Up Call

Hinton considers the possibility that "our best hope is to have AI try to take over and fail." He suggests that something akin to "Chernobyl for AI" or the "Cuban missile crisis" might be necessary to create a sense of urgency and mobilize resources for AI safety. He believes that without such a crisis, "the big companies aren't going to put like a third of their resources into figuring out how to make it safe."