A start-up company called Memvid recently published a controversial recruitment notice, openly recruiting a "professional AI bully" to "torture" mainstream chatbots throughout the day, with a salary of up to $800, sparking discussion among the outside world whether its motivation is a serious technical test or a carefully designed marketing stunt.
Memvid stated that the core purpose of this recruitment is to completely expose to the public the "memory defect" problem that is common in current large models and chatbots. The company believes that although large language models perform well in simulating "meaningful intelligent conversations," they are still essentially deceptive software products with very limited memory capabilities and often "forget" what the user just said in the same round of conversation, causing users to have to reiterate information repeatedly.
According to the job page, this one-day, hourly position is paid at $100 per hour, with total earnings of $800, and is open to remote workers worldwide. Recruits will interact with a number of mainstream AI chatbots in an almost "aggressive" manner at high intensity, continuously requiring them to remember various information. During the process, the system will record situations such as failed conversations, forgotten context, and inconsistencies. In addition to the candidates themselves taking detailed notes, Memvid will also record the entire experiment through screen recording or camera for subsequent analysis.
Interestingly, this job does not require candidates to have any AI technical background or "bullying experience", but it clearly states in the conditions: Applicants need to submit a detailed personal statement explaining their past negative experiences when using technology products, and must have a clear and strong dissatisfaction with current AI technology. Mohamed Omar, the company’s co-founder and CEO, stated that they wanted to find “someone who really hates AI” to complete the task.
Omar pointed out in an interview that the serious memory problems faced by chatbots today are closely related to the "unreliability" of early generation AI memory solutions. In his view, being able to continuously retain user information and maintain contextual coherence in long conversations is the holy grail of "a truly useful chatbot". The so-called "hallucination" phenomenon is largely a direct result of poor large-model memory mechanisms and usage habits.
Memvid takes this opportunity to promote its self-developed AI memory layer solution. The company claims that this is a file-based, self-improving memory component that can provide persistent context across conversations for large models and chatbots, thereby improving response accuracy, reducing latency, and improving infrastructure utilization without sacrificing performance. This memory layer can be used as the core capability of Memvid’s own chatbot product Kora, or it can be integrated into other AI products in the form of developer tools.
As large models are widely deployed in real application scenarios, the problem of AI memory is becoming more and more prominent: in the face of huge data warehouses and complex business requirements, chatbots frequently experience "fragments", forgetting past events, and even making up random things during long-term and multi-round interactions, and the user experience is constantly damaged. In this context, Memvid’s “AI bullying” recruitment is both like a creative advertisement for its own products and a wake-up call to reiterate the current status of the industry, reminding the public that current AI is still far from being a “truly reliable conversation partner.”
Recruitment for the position will continue for several weeks and only one candidate will be selected, Memvid said. The company hopes to obtain enough negative cases through this intensive "torture" of mainstream chatbots, which will not only provide display materials for its own products, but also provide a vivid entry point for social discussions on the safety and reliability of AI.