AGI is Still 30 Years Away — Ege Erdil & Tamay Besiroglu
Apr 17, 2025
auto_awesome
Ege Erdil and Tamay Besiroglu, co-founders of Mechanize and former AI forecasters at Epoch AI, discuss the future of artificial intelligence and automation. They challenge the notion of an imminent intelligence explosion, proposing instead that economic growth will surge over the coming decades. The conversation explores the complexities of AI reasoning, the importance of innovative thinking in research, and the dual pathways of AI and economic development, emphasizing that societal values and human creativity will be vital in shaping this evolution.
Ege Erdil and Tamay Besiroglu predict that advancements in AI will parallel the Industrial Revolution, driving explosive economic growth across multiple sectors.
The timeline for achieving AGI is debated, estimating around 2045, highlighting the complexity and uncertainty surrounding AI development.
Current AI limitations indicate that full remote work automation is still far off, as many tasks require nuanced human input.
The speakers emphasize the need for value alignment in AI systems to mitigate risks and ensure technological benefits enhance human well-being.
Regulatory approaches to AI must balance innovation with safety, as overly restrictive measures could hinder economic progress.
Deep dives
Automation as a New Frontier
The discussion focuses on the launch of Mechanize by Tame Besaroglu and Ege Erdo, which aims to automate all types of work. They draw parallels between their vision for AI automation and the Industrial Revolution, emphasizing that while horsepower accelerated growth, complementary changes across various sectors like agriculture and finance were pivotal. Similarly, they argue that the development of AI will involve simultaneous advancements across many fields, not just in intelligence, which will drive economic organization and technological change. The goal is to highlight a holistic approach to understanding innovation and growth through automation rather than solely focusing on advancements in intelligence.
The Timeline to AGI
The speakers express differing opinions on the timeline for achieving artificial general intelligence (AGI), with estimates around 2045 for one and slightly earlier for another, reflecting both optimism and caution. They believe that current advancements in AI, such as improved familial reasoning capabilities, will accelerate the transition to full remote work automation. However, they also note that many AI systems remain limited in their abilities, indicating that reaching full AGI might take longer than simply extrapolating current progress trends suggests. This complexity in predicting AGI development highlights the necessity of understanding the diverse paths and capabilities inherent in AI advancements.
Economic Scenarios and Computation Scaling
The conversation delves into the computation scaling necessary for AI progress, discussing how advancements require significant compute resources as well as innovative breakthroughs. They assert that the past scaling has unlocked substantial capabilities in AI, but future gains may be constrained by finite compute resources. The speakers discuss the decreasing returns to scale in AI advancements while emphasizing the importance of exploring complementary improvements across various sectors. This perspective poses questions about the sustainability of rapid AI progress and the potential need for new innovations to ensure continuous economic growth.
Challenges of Full Automation
The speakers analyze the possibility of fully automating remote work and highlight that while significant progress has been made, many tasks still require human input to function optimally. They contend that simply achieving a certain level of AI capability does not equate to full automation due to the complexities inherent in various jobs. The duo offers insights into the different competencies AI will need to emulate human workers effectively across tasks that require nuanced understanding and adaptability. This argument illustrates the challenges that remain in achieving broad obsolescence of human roles in the workforce.
Valuation of AI Contributions
The notion of AI's contributions to productivity and economic growth is scrutinized by the speakers, who assert that current assessments may undervalue the expansive potential of AI developments. They pose that while skeptics might highlight the current productivity obstacles, such constraints may diminish as AI systems become better integrated into the labor force. The conversation emphasizes the need to recalibrate our expectations about how AI will generate value and productivity gains across various industries. By reevaluating these contributions, they believe it becomes easier to foresee a world where AI significantly transforms economic landscapes.
The Potential for Economic Disparities
The discussion introduces concerns about the uneven distribution of the benefits arising from AI's growth and the potential for increasing economic disparities among countries. They emphasize that some nations are likely to harness AI technology more efficiently, leading to varying rates of economic expansion. This theoretical framework invokes historical parallels to the Industrial Revolution, where some countries advanced more rapidly, leaving others behind. The speakers state that real-world implementation dynamics and regulatory environments could influence how these disparities unfold, highlighting the importance of nuanced policy approaches.
Aligning AI with Human Values
The speakers discuss the importance of ensuring that AI systems are aligned with human values, considering both ethical and practical implications. They argue that as AI technologies progress, a greater emphasis on values must be embedded in their design and implementation. This intention could mitigate potential risks associated with misaligned AI behaviors, as well as forecast the broader implications of AI on society. An adequate focus on value alignment will help shape a future where AI technologies support rather than undermine human well-being.
Regulatory Considerations
The challenge of regulation in the context of AI advancements is a significant focus of the conversation, with the potential for regulatory frameworks to hinder progress being stressed. The speakers highlight that while regulation can be essential for addressing concerns related to safety and ethical use, overly restrictive measures may stifle innovation and economic growth. They argue that a careful balance must be struck to facilitate AI's integration into wider society while ensuring robust safeguards are in place. This commentary emphasizes the need for adaptive regulatory approaches that evolve alongside rapidly advancing AI technologies.
Collective Intelligence of AI Systems
The conversation addresses the idea of collective intelligence among AI systems, which can create unprecedented efficiencies not possible with human workers. The speakers suggest that AI systems can be replicated and integrated in ways that enable them to collectively solve complex challenges faster and more effectively than humans. This paradigm shift poses significant implications for the structure of organizations and the nature of work, suggesting a potential redefined workforce as human roles evolve. They argue that harnessing this collective intelligence could lead to substantial advancements across a plethora of sectors.
Navigating Uncertainty in AI's Future
The speakers underscore the importance of embracing uncertainty when navigating the future of AI and the broader technological landscape. They advocate for sustainable approaches to policy, research, and collaboration, as certainty often becomes elusive in rapidly changing environments. Instead of rigid stances, they stress the value of open channels for dialogue and a decentralized approach to innovation that can better adapt to emerging needs and challenges. This perspective promotes the idea that flexibility and adaptability will ultimately lead to more resilient outcomes as AI technologies continue to evolve.
Ege Erdil and Tamay Besiroglu have 2045+ timelines, think the whole "alignment" framing is wrong, don't think an intelligence explosion is plausible, but are convinced we'll see explosive economic growth (economy literally doubling every year or two).
Ege and Tamay are the co-founders of Mechanize (disclosure - I’m an angel investor), a startup dedicated to fully automating work. Before founding Mechanize, Ege and Tamay worked on AI forecasts at Epoch AI.
* WorkOS makes it easy to become enterprise-ready. With simple APIs for essential enterprise features like SSO and SCIM, WorkOS helps companies like Vercel, Plaid, and OpenAI meet the requirements of their biggest customers. To learn more about how they can help you do the same, visit workos.com
* Scale’s Data Foundry gives major AI labs access to high-quality data to fuel post-training, including advanced reasoning capabilities. If you’re an AI researcher or engineer, learn about how Scale’s Data Foundry and research lab, SEAL, can help you go beyond the current frontier at scale.com/dwarkesh
