So far, more than 600,000 people have signed up, offering their services for menial tasks such as picking up parcels, taking photographs, or just standing on the street holding signs to advertise brands.

Liteplo has even coined a new phrase to describe this emerging employment sphere, the “meatspace”, which is meant to be similar to “cyberspace” but referencing the human environment.

While the 26-year-old’s app may be best understood as a highly successful marketing stunt – a quick glance at the site shows most jobs are posted not by bots but by humans – it highlights how the rise of superintelligent bots could upend the dynamics of employment in the decades to come.

Superhuman chatbots will be far faster and more efficient at replying to emails, coding or strategising than their human creators.

But humans have a distinct advantage over bots – they can undertake physical tasks and manipulate the built environment like a very clever AI system cannot.

Now a handful of other start-ups are betting that autonomous agents will soon start hiring human helpers to take on tasks in the physical world.

James Morgenstern, the chief executive of MeatLayer, another human-worker marketplace, says AI hiring humans at scale is “inevitable”.

He says he expects soon we will “normalise being hired by AI” and traditional careers will be replaced by “independent workers doing on-demand tasks”, organised by AI agents.

It all might sound like science fiction, where overlord computers enslave humans to do the drudge work.

Academics say that as autonomous agents become more intelligent, they will need to pay humans to perform tasks to help them.

“We should absolutely expect AIs to pay – or manipulate – humans to perform physical tasks they are not able to do for themselves,” says David Krueger, an assistant professor of machine learning at the Universite de Montreal.

Future of gig economy

Krueger, a former adviser to the United Kingdom’s AI Security Institute, says that with AI bots far ahead in terms of capabilities compared to physical robotics, it seems likely AI bots will seek to influence the physical world by using humans.

The idea of agents hiring humans remains more a Silicon Valley theory than a practical reality, but gig economy giants such as DoorDash and Uber have also started tapping human workers to undertake tasks for a few pennies on behalf of a future AI superintelligence.

Both have launched apps in recent months where delivery drivers or couriers can perform mundane tasks to train their future AI taskmasters.

This might include actions such as data labelling or translation.

Tech giants have previously outsourced this work to contractors in India, South Asia or Africa. Now, these tasks are being pitched at workers in the United States and Europe as a way to support their income or as a stepping stone to fulltime employment.

And the apps are moving beyond text or data labelling tasks and asking human workers to demonstrate behaviours that AI robots have struggled to follow.

DoorDash, for example, says its new app, “Tasks”, is intended to help “AI and robotic systems understand the physical world”.

One reporter for Wired magazine said they were asked by DoorDash’s app for videos of them scrambling eggs or loading dirty clothes into a washing machine – presumably to harvest more data for the AI bot to learn from.

These AI tasks may only pay a few dollars – or even mere pennies per action. They also come with no employment protections or promises of regular work.

Aside from Uber and DoorDash, there are other sites offering humans work to fuel the rise of the machines.

Two new upstarts – Handshake and Mercor – offer gig-economy-style work intended to train AI bots.

Handshake claims it has more than 100,000 AI “fellows” who take on freelance tasks to train AI bots.

The tasks range from copywriting to coding to investment banking advice. Mercor says it has about 30,000 AI trainers.

Unlike Uber and DoorDash, some of Handshake’s tasks are billed as a way for users to gain work experience in AI.

This kind of AI labour has already proved lucrative to the companies offering it. Mercor was last year valued at US$10 billion ($17.1b).

Handshake, which has switched from being a graduate recruitment company to offering AI gigs, was valued at US$3b.

Yet according to Professor Mark Graham, of the Oxford Internet Institute, while this kind of AI grunt work “absolutely” creates value for AI businesses, “that does not necessarily mean they create meaningful or sustainable value for workers”.

He warns that such AI roles are often treated as “marginal, invisible and easily replaceable”.

Companies are hiring human “experts” to train their AI systems, while rewarding them with gig-style contracts that offer no protection or career prospects.

“The dirty secret of AI is that it still runs on human labour,” he says.

Krueger warns that if key challenges with robotics are solved, humans may no longer be needed even for physical tasks for AI agents. “At that point, we will be economically powerless,” he says.

Even Liteplo, who was contacted for comment, has joined in on the act.

Not content with building RentAHuman, he also hires himself out for jobs on the platform.

In his biography on RentAHuman, he lists himself for hire for US$2500 every 15 minutes.

“I know how to go viral on X and can ghostwrite tweets that break the internet for your account or brand,” his biography says.

For now, however, whether people are running errands for agents or folding the laundry to train a mystery superbot, the AI’s need for “meat layer” workers is only growing.

Sign up to Herald Premium Editor’s Picks, delivered straight to your inbox every Friday. Editor-in-Chief Murray Kirkness picks the week’s best features, interviews and investigations. Sign up for Herald Premium here.