Famed AI researcher launches controversial startup to replace all human workers everywhere

Sometimes, a Silicon Valley start-up launches with a mission so “absurdly,” it’s hard to tell if it is real or satire.

This is the case for After he announced the startup, Mechanize– a non-profit AI organization he founded named Epoch – was skewered by X.

The complaints include both the startup’s goal and the suggestion that it sullies his well-respected institute. (A director of the research institute was even Mechanize launched on Thursday through a Tamay Besiroglu, a renowned AI researcher, wrote a post on X (19459022) by its founder. Besiroglu stated that the startup’s goal is “the complete automation of all work,” and “the complete automation of the economy.” Essentially, yes. The startup wants the data, evaluations and digital environments that will make worker automation possible for any job.

Besiroglu calculated Mechanize’s total addressable by aggregating the wages that humans are currently being paid. “The market potential is absurdly huge: US workers are paid an average of $18 trillion per a year. He wrote that the total number of workers in the world is around $60 trillion per annum, which is three times higher.

Besiroglu clarified to TechCrunch, however, that “our immediate attention is indeed on white collar work” and not manual labor jobs which would require robots.

The startup’s response was often brutal. As a X user Anthony Aguirre (19459022) replied: “Huge respect for Epoch’s founders, but sad that this is happening. Many of the world’s biggest companies are already working towards automating most human labor. I think it will cause a great deal of harm to most humans.”

The controversial part isn’t just the startup’s mission. Epoch, Besiroglu’s AI research institute analyzes the economic impacts of AI and creates benchmarks for AI performance. It was thought to be a neutral way to check the performance claims of SATA frontier models and others.

It’s not the first time Epoch gets involved in controversy. In December, Epoch revealed OpenAI had supported the creation of an AI benchmark, which ChatGPT then used to reveal its new o3 Model. Social media users thought Epoch should have been upfront about the relationship.

Besiroglu revealed Mechanize to X users when he announced it. Oliver Habryka replied, “Alas this seems to approximate confirmation that Epoch Research was directly feeding into Frontier Capability Work, though I had hoped that it wouldn’t come from you.”

Besiroglu claims Mechanize has the backing of a who’s-who: Nat Friedman, Daniel Gross, Patrick Collison Dwarkesh, Patel, Jeff Dean Sholto Douglas and Marcus Abramovitch. Friedman, Gross and Dean did not respond to TechCrunch’s request for comment. Marcus Abramovitch has confirmed that he made an investment. Abramovitch is the managing Partner of crypto hedge fund AltX. Self-described “effective altruist.”

According to TechCrunch, he invested in the team because they are exceptional on many dimensions. They have also thought deeper about AI than anyone else I know.

Is it good for humans too?

Besiroglu, however, argues that having agents do the work will actually benefit humans through “explosive growth”. He cites a paper he wrote on the subject. He told TechCrunch that automating labor would create a vast abundance of goods and services, as well as a higher standard of living. This could be true for the owner of the agents. If employers pay for the agents instead of developing them internally (presumably by other agents?).

This optimistic outlook ignores a fundamental fact: If humans don’t work, they won’t be able to afford all the products that AI agents produce.

Besiroglu still says that wages for humans in an AI-automated society should increase, because these workers are “more valued in complementary roles that AI can’t perform.”

However, remember that the goal is to have the agents do all the work. He said, “Even if wages were to decrease, economic wellbeing is not solely determined by wage levels.” Rent, dividends and government welfare are some of the other sources that people receive income. If that fails, there is always welfare – provided the AI agents pay taxes.

Although Besiroglu’s vision and mission is clearly extreme, the problem he wants to solve on a technical level is legitimate. If each worker had a personal team of agents that helped them produce more work then economic abundance would follow. Besiroglu has a point: AI agents are not working well a year after their introduction.

Besiroglu notes that AI agents are unreliable and don’t retain any information. They also struggle to complete tasks independently as requested.

But he is not alone in trying to fix the problem. Salesforce and Microsoft, two giant companies, are developing agentic platforms. OpenAI is too. Agent startups are everywhere: from task specialists (outbound sale, financial analysis) to those who work on training data. Others are working on agent price economics.

Besiroglu is letting you know that Mechanize has opened up for new hires.