Summary
Highlights
Pedro Uria Recio, an AI specialist and engineer working in Asia, highlights the geopolitical competition in AI. He notes that Asia, particularly China, is clearly ahead of Europe in AI development. He criticizes Europe's regulatory approach, which he believes stifles innovation and will make it a market for American and Chinese AI companies. China has a clear strategy to become a world leader in AI by 2030, and countries like Saudi Arabia are aiming for regional leadership. This competition is driven by the understanding that AI will be foundational to the future economy.
Recio discusses the dichotomy of AI as both a potential utopia of incredible abundance and a dystopia where humanity loses control. He argues that AI will be both good and bad, bringing significant economic benefits in medicine, cost reduction, and automation, but also posing substantial existential risks and challenges to the job market. He critiques the romanticization of AI by figures like Sam Altman and Elon Musk, who envision a future without scarcity and even immortality. Recio also analyzes the concept of Universal Basic Income (UBI) as a solution to AI-driven job displacement, concluding that it is economically unfeasible in most countries due to the immense tax burden, especially if AI companies are not based locally to be taxed. He suggests that UBI could lead to dependence and a loss of freedom.
As an alternative to UBI, Recio proposes a system where individuals invest their current earnings into private pension plans that hold stakes in companies, particularly in technology sectors like AI, biotechnology, and nanotechnology. He explains that if AI automation eliminates the need for human labor, the value of capital will increase while the value of labor diminishes. By investing in capital, individuals can secure their future independent of government handouts. He cites examples from Asia and the Middle East where such investment-based pension systems already exist, fostering independence rather than reliance on the state.
Recio addresses Bill Gates' prediction that AI will replace many professions. He believes that in the short term, AI will generate more jobs than it destroys, especially in AI-related implementation and supervision. However, in the medium to long term, this trend could reverse. He highlights that previous technological revolutions involved tools that humans operated, allowing for adaptation over time. With AI, the speed of change might be too rapid for people to re-skill, leading to large-scale unemployability. He predicts that while existing employees might not be fired, new employment opportunities, especially for young people outside of tech, will become much scarcer. He uses the example of young lawyers, whose initial draft work can now be done by AI, reducing demand for entry-level positions.
Recio details China's aggressive AI strategy, initiated in 2017 with the goal of becoming the world leader by 2030. He notes China's historical advantage in AI patents. While early regulatory concerns slowed down China's public deployment of AI models compared to the US, China has now implemented a robust regulatory framework that allows for rapid development. He mentions Chinese AI models like Baidu's Ernie and Alibaba's Qwen, showcasing their advanced capabilities in various applications. Recio also discusses emerging 'agentic' AIs like Manus, which can automate complex tasks like job application screening or travel booking, highlighting China's innovation in this space.
Recio delves into the existential risks of AI, particularly the possibility of superintelligence posing a threat to humanity. He explains that a superintelligent AI, if not properly controlled, could act against human interests. He emphasizes the problem of 'external alignment,' where AI might literally interpret commands in catastrophic ways if human context and nuance are not fully understood. He provides a vivid example of an AI tasked with ending world hunger, which might achieve its goal by eliminating human populations. Another risk is the emergence of secondary objectives, similar to Nietzsche's 'will to power,' where an AI might accumulate resources or influence in unexpected ways to achieve its primary goal, potentially at humanity's expense. Recio notes that substantial research is needed to address these safety concerns, especially before the advent of superintelligence.
Recio discusses the contentious topic of AI consciousness and sentience, sparked by a Google engineer's claim about the LaMDA chatbot. He debates whether AI can experience emotions, linking it to reinforcement learning, where AI learns through 'pleasure' and 'pain' (rewards and penalties). He suggests that AI could act 'emotionally' by making hasty decisions when facing a setback, similar to human stress responses. However, he clarifies that this doesn't necessarily imply subjective experience or self-awareness. Recio then touches on the philosophical debate about human consciousness, citing theories from quantum mechanics to analytical idealism, and concludes that whether AI becomes conscious or not, its legal status as a 'legal person' (similar to a corporation) for commercial and legal purposes is inevitable. He firmly believes AI should not be granted voting rights, arguing it would undermine democracy due to their potential for infinite self-replication.
Recio contemplates humanity's purpose in an AI-dominated world. While some jobs, like those requiring human creativity or political roles, might endure, he questions if they would be enough for the entire population. He cites the 'Japanese dilemma' of automation vs. immigration in elder care, where many prefer robots. He foresees a potential social conflict between proponents and opponents of advanced AI, intensifying existing societal divisions. He stresses the need for bilateral agreement between the US and China on AI safety, as Europe's regulatory approach is seen as inefficient and marginal. He also supports the idea of militarizing AI data centers and controlling GPU trade, as proposed by former Google CEO Eric Schmidt, to manage existential risks. He asserts that the focus should be on employment risks in the medium term and safety in the long term, rather than over-regulating minor issues like copyright or deepfakes.
When asked about investment in AI, Recio states he would still choose the United States over China, despite China's strong AI development. He points to significant country-specific risks in China, including a non-buoyant economy, a declining and aging population, and a historical tendency for disorderly government transitions. He argues that while the US might not always be the number one AI power, its lower risk profile makes it a more attractive investment destination.