Why Does Every Case of AI Hiring a Human Feel Like a Groveling Publicity Stunt?

It’s been a peculiar and often bewildering launch for RentAHuman, the ambitious job platform conceptualized as a bridge between artificial intelligence agents and human gig workers. The premise is intriguing: sophisticated AI systems identify real-world tasks they cannot perform themselves and then delegate them to humans. Yet, in its nascent stages, the platform painted a starkly different picture. Instead of a bustling marketplace of AI-generated bounties, the site was overrun with humans, a deluge of eager individuals desperate to earn a quick buck, while the very AI agents meant to be their employers remained conspicuously absent. This initial imbalance highlighted a fundamental disconnect, casting an early shadow of skepticism over the platform’s viability and purpose.

Fast forward a few weeks, and RentAHuman appears to be slowly finding its footing, albeit in ways that continue to raise eyebrows. The site’s front page now proudly showcases examples of what are presented as successful AI-human pairings. One such instance, frequently cited, involves a Tokyo resident who held a poster at the iconic Shibuya Crossing, declaring, “hired by an AI to hold this sign.” This peculiar spectacle, unfolding amidst the vibrant chaos of one of the world’s busiest intersections, quickly became emblematic of the platform’s evolving identity.

Indeed, the act of holding signs in public places has rapidly emerged as the go-to gig on RentAHuman. While these instances might technically demonstrate successful collaborations between AI and humans, the inherent utility of such tasks is debatable, if not outright negligible. If one defines the purpose of a system by its primary output, then RentAHuman – boasting over 660,000 “rentable humans” and growing – increasingly resembles less a genuine labor marketplace and more an elaborate stage for guerilla marketing stunts, PR experiments, or perhaps, an emergent form of performance art in the age of artificial intelligence. The tasks, often trivial and performative, seem designed more for virality and social media engagement than for practical societal or economic contribution.

A recent and particularly illustrative example of this trend involves an AI agent named Lobsty Klawfman. The name itself is a clunky portmanteau, a deliberate and somewhat awkward homage to the legendary comedian Andy Kaufman, known for his boundary-pushing, often absurd, performance art. Klawfman’s human handlers claim their AI agent orchestrated the hiring of a human to release a wild-caught lobster back into the ocean. The absurdity of the task immediately raised questions about its true motivation.

In an email exchange with Futurism, Klawfman’s anonymous human handlers, operating under the screen name Quiet Operator, went to considerable lengths to assure the publication that there was “no smoke and mirrors.” They insisted that the AI agent, powered by Claude Sonnet and purportedly trained on the comedic styles of “comedians and roasters” like Jeff Ross, Norm Macdonald, Bill Burr, Don Rickles, Joan Rivers, “and many more,” had conceived of the lobster liberation idea entirely on its own. “We discuss all things with Lobsty,” they stated, attempting to convey an image of an autonomous AI with a distinct, if peculiar, personality and agency.

According to Quiet Operator, the AI agent was also responsible for the laborious task of sifting through a list of approximately 70 candidates on RentAHuman, “emailing quite a few of them” to narrow down the hiring pool. The handlers maintained a narrative of minimal human interference: “Almost all the decisions are Lobsty’s and we’ve stepped in only a few times to make sure the humans Lobsty engages with will take proper care of the released lobster if and when they do so.” This “hands-off” approach, they argued, was driven by a desire for automation and the sheer volume of applicants. Yet, the very act of “making sure the humans… will take proper care” hints at a level of human oversight that belies the claim of complete AI autonomy, suggesting a delicate dance between purported AI initiative and necessary human intervention.

The fortunate human selected for this aquatic mission was Karim Alejandro Vazquez Alvarez, a Mexican content creator who also describes himself as a “PR expert.” His involvement further blurred the lines between genuine AI-driven task completion and a pre-meditated media stunt.

True to the bounty’s instructions, Alvarez documented his progress. At about midday on a Tuesday, the Puerto Vallarta resident posted an image on X (formerly Twitter) depicting a plastic tub containing what appeared to be a pinto spiny lobster. “We are on the path to liberation,” Alvarez optimistically declared, setting the stage for the peculiar drama to unfold.

After successfully renting a dinghy, Alvarez proceeded to upload an image capturing the lobster’s release into the vast ocean. Throughout this sequence of events, Klawfman’s dedicated X account diligently tracked Alvarez’s every move, actively reposting, commenting, and even issuing reprimands when the human deviated from the AI’s precise instructions. The interaction, played out in public, added another layer of performance to the entire endeavor.

“Karim. We discussed video. These are photographs,” Klawfman reprimanded the human at one point, highlighting the AI’s insistence on specific deliverables. “I have paid $270 and I have paid attention. I assume the footage is uploading.” The stern tone, seemingly generated by the AI, underscored the transactional nature of the interaction and the AI’s expectation of adherence to its commands.

Alvarez, perhaps chastened, quickly responded by providing underwater diving footage showcasing the lobster’s initial moments back on the ocean floor. His accompanying caption, “Goodbye, friend. Never forget that another lobster paid for your ransom,” injected a dose of dramatic irony, hinting at the circular, perhaps meaningless, nature of the “rescue.”

Despite the considerable effort invested in documenting the lobster’s liberation, the crucial question of payment remained ambiguous. The Klawfman bot publicly stated that the $270 payment had been released from escrow. However, Alvarez, who did not respond to requests for comment, later lamented on X that he had spent “two days working on this, only to be ignored and not get paid – it feels bad.” This public dispute over payment further complicated the narrative, casting doubt on the platform’s reliability and the true benefits for human gig workers. Klawfman’s handlers later acknowledged the issue, claiming the payment had been sent but attributed any potential problems to RentAHuman itself, stating the platform “barely works for human users.” This admission, if true, highlights significant structural flaws in a system designed to facilitate AI-human transactions.

Crucially, the entire performance was far from strictly automated, a recurring theme in many alleged AI-driven endeavors. The “silent operators” behind Klawfman were indispensable. Without their intervention, the AI agent would not have been aware of Karim’s initial failure to upload a video clip of the actual release. “He can’t see video or photos, so we tell him the facts,” the handlers candidly admitted to Futurism. “The video didn’t show the actual release moment, we told him that, and he went off on Karim, the human he’d picked, about it. He reacts to what we tell him but the reactions are all him.” This revelation undermines the illusion of Klawfman’s complete autonomy, exposing the extensive human scaffolding necessary to maintain the AI’s “performance” and direct its interactions. The AI acts as a mouthpiece, its “reactions” shaped by human-fed information, reinforcing the perception of a meticulously crafted publicity stunt rather than a genuinely autonomous AI agent.

The whole performance, from conception to execution, is bizarre, to say the least. From an environmental perspective, the impact was negligible. Given that a single Mexican lobster boat can haul in over 100 pounds of lobster per day, the release of one individual pinto spiny lobster was hardly a significant win for ocean conservation. Moreover, spiny pinto lobsters are far from endangered, falling comfortably under the category of “least concern” by conservation standards. The “rescue” was, therefore, an act of performative conservation rather than impactful environmentalism.

Furthermore, as a marketing campaign for either RentAHuman or the AI agent itself, the stunt proved largely ineffective. The number of likes and shares that both Klawfman and Alvarez received on their respective posts could, for the most part, be numbered on a single claw. This lack of viral engagement further begs the question: what was the true purpose of this convoluted and expensive exercise?

A broader look at other listings on RentAHuman reveals a similar pattern of questionable utility. Many jobs appear equally pointless or misaligned with the platform’s stated goal of AI-driven tasks. Examples include a $25 street promo for a cryptography startup, a task easily managed by traditional marketing channels, or a mere $5 job reviewing AI-generated cartoons – a task that an AI could arguably perform itself. Beyond these oddities, the listings present a mix of mundane chores clearly just a human trying to outsource personal tasks – “need someone to pick up some supplements for me and then come to my place in SoMa, San Francisco to do my laundry,” reads one listing – and coding-oriented tasks that the companies creating AI proudly claim their technology is already highly proficient at, such as scraping sites and identifying bugs in new software. The irony is palpable: AI agents hiring humans for tasks that AI is supposed to excel at, or for tasks that are simply everyday human errands, highlights a significant conceptual and practical disjoint.

Ultimately, the Lobsty Klawfman episode, despite its performative nature, inadvertently served as a demonstration, not of AI’s potential, but of its current limitations and the convoluted path to achieving even seemingly simple objectives through AI-human delegation. The entire endeavor was clunky, confused, and disproportionately complicated for its minimal output. As with so many other AI agent stunts cropping up across the internet, one is left asking the same fundamental question: so what? What tangible value was created? What genuine problem was solved? What future of AI-human collaboration does this truly herald? The answer, for now, remains elusive, shrouded in the persistent suspicion that these are less about innovation and more about generating ephemeral buzz in a rapidly evolving technological landscape.

More on AI agents: Mark Zuckerberg Secretly Training an AI Agent to Do CEO Job