Contents
The Vision of Automating Every Job with AI
Tamay Besiroglu, cofounder and CEO of Mechanize, has a bold vision: to automate every job using artificial intelligence. His company is starting with software engineering, a field that involves more than just writing code. Besiroglu believes that while AI can assist with many tasks, fully replacing human software engineers could take a decade or more.
“Software engineering is just a lot more than writing code,” Besiroglu said. Engineers need to understand their employer’s goals and communicate with product managers and customers. This complexity makes the task of automation more challenging than it might appear at first glance.
Mechanize is using a technique called reinforcement learning (RL) to train AI. In this process, an AI agent is rewarded for completing tasks correctly. The company builds these environments to sell to AI labs, which use them to train their models. This approach aims to scale up AI capabilities, making it possible for AI to handle more complex tasks over time.
Besiroglu, who previously founded the AI nonprofit research organization Epoch AI, started Mechanize this year with Matthew Barnett and Ege Erdil. The startup has backing from notable figures like Jeff Dean, Google’s chief scientist, and Stripe CEO Patrick Collison. Despite the controversy surrounding its mission to replace all jobs with AI, Besiroglu sees the discussion as important and necessary.
Why Start with Software Engineering?
There are several reasons why Mechanize chose to focus on software engineering first. One key factor is the existing culture of testing in software development. These tests help determine if a piece of software works as intended, which is crucial for reinforcement learning. In RL, an AI needs a clear reward signal to know if it performed a task correctly.
Another reason is the competition among tech giants like Google, OpenAI, and Anthropic. These companies are investing heavily in coding, making it easier for Mechanize to secure lucrative deals. This competitive landscape provides opportunities for the startup to grow and develop its technology.
The Transition to an AI-Driven Economy
Besiroglu acknowledges that the transition to an AI-driven economy will be gradual. While some industry experts predict rapid changes within the next decade, he believes it will take decades to develop the necessary technology and infrastructure. This slower pace allows society to adapt and find solutions to potential challenges.
He also points to examples like Norway, where a sovereign wealth fund supports citizens even without traditional labor income. Such models could provide insights into how societies might adjust to widespread automation. While environmental concerns are valid, Besiroglu believes AI can also drive technological progress that helps mitigate these issues.
The Future of Work and Robotics
Despite focusing on remote jobs, Besiroglu is not ignoring the future of robotics. He notes that robotics is more complex and will take longer to develop. However, he remains optimistic about advancements in humanoid robotics, envisioning a future where robots can handle household tasks.
The irony of hiring engineers to eventually replace themselves is not lost on him. Yet, he emphasizes that these engineers are working on significant problems and are rewarded well, including through equity. This model ensures that even as AI evolves, those involved in its development can still benefit financially.
Conclusion
As Mechanize continues to push the boundaries of AI, the journey toward full automation remains complex and multifaceted. The company’s focus on software engineering is just the beginning, with broader implications for the future of work and economic growth. While challenges remain, the potential benefits of AI-driven automation are vast, offering new opportunities for innovation and societal advancement.
