AI startup Mechanize seeks to fully automate work
CALIFORNIA, UNITED STATES — Mechanize, a new artificial intelligence startup based in San Francisco, has set an ambitious goal: to “fully automate work,” according to co-founder Tamay Besiroglu.
The company is starting with software engineering, using reinforcement learning to train AI systems to perform tasks typically handled by humans using everyday workplace apps.
“Our goal is to fully automate work,” Besiroglu told The New York Times. “We want to get to a fully automated economy, and make that happen as fast as possible.”
Reinforcement learning at the core of Mechanize’s technology
Mechanize’s approach centers on reinforcement learning, a technique previously used to train AI to master games like Go.
The company has built virtual environments that mimic a software engineer’s workstation, complete with email, Slack, coding tools, and a web browser.
The AI is tasked with completing assignments in this simulated setting. Success is rewarded, while failure results in penalties, allowing the AI to improve through trial and error.
“It’s effectively like creating a very boring video game,” Besiroglu explained. The company believes that, with enough repetition, AI will learn to perform the same functions as a human engineer.
Broader ambitions and industry impact
While Mechanize is beginning with programming, its founders hope the same strategy can eventually automate a wide range of white-collar jobs.
The company, founded this year by Besiroglu, Ege Erdil, and Matthew Barnett, has already attracted investment from tech leaders such as Stripe’s Patrick Collison and Google’s Jeff Dean.
Mechanize currently employs five people and is working with leading AI companies, though it has not disclosed which ones.
Despite the bold vision, the founders acknowledge the challenges ahead. Barnett estimates that achieving full automation could take 10 to 20 years, while Erdil and Besiroglu expect it may require 20 to 30 years.
Economic and social implications
Mechanize’s candid approach stands out in an industry where many companies are less open about their automation ambitions. However, some observers express concern about the social impact of such rapid change.
While Besiroglu envisions a future of “radical abundance” and supports ideas like universal basic income, the company offers no concrete policy proposals to address potential job displacement.
“If society as a whole becomes much wealthier, then I think that just outweighs the downsides of people losing their jobs,” Barnett said, emphasizing the team’s belief in the long-term benefits of automation.
