Sometimes Silicon Valley startups launch on a “absurd” described mission that it is difficult to identify whether a startup is a real or mere satire.
This is the case with Mechanize, a startup that has been skewered in X after the founder, and the non-profit AI research organization he calls Epoch.
The complaint covers both the startup's mission and the meaning that it makes the reputation of his respected institute unhappy. (The Director of Research Institute was posted on X.
Mechanize was released on Thursday via X's post by renowned AI researcher Tamay Besiroglu. The goals of Startup are “Full automation of all tasks” and “Full automation of the economy.”
Does that mean mechanization is working to replace all human workers with AI agent bots? Basically, yes. Startups want to provide data, assessment and digital environments to enable automation of workers at any job.
Besiroglu calculated Mechanize's total addressable market by aggregating all wages that humans are currently paid. “The market possibilities here are ridiculous. U.S. workers get paid a total of about $18 trillion a year. Globally, the numbers are more than tripled, about $60 trillion a year,” he writes.
But Besiroglu made it clear to TechCrunch that “our immediate focus is certainly on white-collar work,” rather than manual labor work that requires robotics.
The reaction to startups was often brutal. As X user Anthony Aguirre answered, “It's a great respect for the founder's work at Epoch, but it's sad to see this. Automation of labor for most people is actually a big prize for businesses. So, many of the biggest companies on the planet have already pursued it. I think it's a huge loss for most people.”
But the startup's mission isn't the only controversial part. Epoch, Besiroglu's AI laboratory, analyzes the economic impact of AI and generates benchmarks for AI performance. It was considered a fair way to confirm performance claims from SATA Frontier Model Makers and others.
This is not the first time Epoch has been caught up in a controversy. In December, Epoch revealed that Openai supports creating one of its AI benchmarks. This was used by ChatGpt-Maker to unveil the new O3 model. Social media users felt that Epoch should have had more advances on the relationship.
When Besiroglu announced mechanization, X user Oliver Habryka said, “Although this seems like an approximate confirmation that epoch research is directly fed into frontier capabilities work, I literally hoped it wouldn't come from you.”
Besiroglu says mechanization is supported by whom WHO: Nat Friedman and Daniel Gross, Patrick Collison, Dwarkesh Patel, Jeff Dean, Sholto Douglas, Marcus Abramovitch. Friedman, Gross and Dean did not reply to TechCrunch's request for comment.
Marcus Abramovitch confirmed that he had invested. Abramovitch is a managing partner of Crypto Hedge Fund Altx and is a self-proclaimed “effective altruist.”
He said he invested in TechCrunch.
Is it good for humans too?
Still, through “explosive economic growth,” Besiroglu insists to deniers that if agents do all their jobs, it doesn't actually make them poor, but actually make them rich. He refers to a paper published on this topic.
“Fully automating labor can create enormous amounts, much higher standard of living, and new products and services that you can't even imagine today,” he told TechCrunch.
This may be true for those who own an agent. That is, if the employer pays for them instead of developing in-house (perhaps by other agents?).
Meanwhile, this optimistic outlook overlooks basic facts. If humans don't have jobs, they don't have the income to buy everything that AI agents are generating.
Still, Besiroglu says that the wages of humans in such an Ai-Automated world should actually increase because such workers are “more valuable in complementary roles that AI cannot perform.”
But remember, the goal is for the agent to do all the work. When asked about it, he explained: “Even in scenarios where wages may decrease, economic well-being is not determined solely by wages. People usually receive income from other sources, such as rent, dividends, government welfare.”
So perhaps we all make our living from stocks and real estate. If you fail, there is always welfare if your AI agent pays taxes.
Besiroglu's vision and mission are obviously extreme, but the technical issues he is trying to solve are legal. Financial abundance can last if each human worker has a personal crew of agents that will help create more work. And Besiroglu is definitely the right thing about at least one thing. A year after the age of AI agents, they don't work very well.
He should not keep information, he should not struggle to complete tasks independently as they were asked.
But he's hardly just working on fixes. Huge companies like Salesforce and Microsoft are building agent platforms. So does Openai. There are a lot of agent startups: from task experts (outbound sales, financial analysis). For those working on training data. Others are working on agent pricing economics.
In the meantime, Besiroglu wants you to know: mechanization is employment.