Emad Mostaque: AI Agents Just Broke Out On Their Own

Emad Mostaque argues that the societal impact of AI will surpass that of COVID, with AI agents becoming increasingly human-like in their interactions. He predicts that the next generation of AI models will be 10 to 100 times faster and significantly cheaper, with current models expected to be a thousand times less expensive by next year.

Mostaque emphasizes the importance of AI ownership being in the hands of the public, foreseeing a future where robots dominate the economy. He points out that robots do not require traditional currency but rather compute and electricity, while also acknowledging the potential for malevolent uses of AI as its intelligence grows. He warns that biases in AI models, stemming from the backgrounds of data labelers, could lead to unequal treatment.

He expresses concern that outsourcing cognitive tasks to AI may diminish human intelligence and suggests that the U.S. government might impose requirements for AI models to exhibit patriotic behavior, which could influence their outputs. Mostaque highlights the necessity of private inference to safeguard user data and sovereignty, alongside the rising interest in NFTs due to their inherent scarcity.

Mostaque notes that the U.S. aims to lead in AI agents to boost the economy, with the next generation of cryptocurrency focusing on coordination and value transfer. He reveals plans for an intelligence company in New York that aims to provide every resident with an AI agent and a wallet, allocating equity to children under 20 and newborns in specific regions. He stresses the need for intelligence to be distributed and controlled by the populace while warning of potential infrastructure failures in the near future.

He points out that the average person struggles to effectively prompt advanced AI models, indicating a gap in user capability. Mostaque envisions a future where swarms of AI agents play a vital role in the intelligent internet, outlining a strategic plan that includes forming a founding council and securing capital for this initiative.

Mostaque advocates for a decentralized approach to AI development, suggesting that local champions will be crucial for the last mile of intelligence. However, he cautions that time is running out for grassroots AI development, emphasizing the urgent need for institutional involvement. He notes the growing public concern over accountability in the development and management of AI technologies that affect daily life.

Looking ahead, Mostaque predicts that within five to ten years, the goal is to establish state champions owned by the people, reflecting a shift towards community-driven governance of AI.