OpenAI will pay ₹3.7 crore to worry about a problem that may not exist yet. The job? Stop AI from building itself
OpenAI has posted a job posting for a researcher who will work on one of AI’s most important challenges: What happens when an AI system has the ability to make itself smarter? Located on the company’s Prep security team, the position has a salary range of $295,000 to $445,000 (approximately). ₹2.5 crore ₹3.7 crore) and is an unusual hiring criteria, according to Business Insider, which first reported the list.
“This work is based on reasoning about problems that may exist in the future but may not exist now,” the listing reads, according to Business Insider. “So it is especially important for people in this role to be tasteful and strategic.”
What is Recursive Personal Development and Why is it Important?
The OpenAI backlog focuses on a concept known as recursive self-improvement, the ability of an AI system to research, design and train better versions of itself without meaningful human involvement. The idea has moved from theoretical concern to active industry priority in the last six months, as OpenAI and Anthropic’s coding tools advance at a pace that surprises even their own researchers.
Google DeepMind general manager Demis Hassabis said this week that humanity is now “on the outskirts of the singularity,” at the point where artificial intelligence begins to improve itself and surpasses human intelligence.
Researchers at METR, a laboratory that studies AI model capabilities, wrote in March that the length of a task that leading AI models can complete is doubling approximately every seven months. What this means, METR writes, is that AI agents will soon be able to handle “a large portion” of the software work that human coders take days or weeks to complete.
OpenAI’s Own Timeline for Automated AI Research
OpenAI CEO Sam Altman clearly stated the company’s goals in this area. In October, Altman said the company set a goal of employing hundreds of thousands of on-chip “automated AI research interns” by next September and “true automated AI researchers” by March 2028.
“We may fail completely in this goal, but given the extraordinary potential impacts, we think it is in the public interest to be transparent about it,” Altman wrote of X.
The company, which aims to go public this year, is already commercializing artificial intelligence coding tools through its Codex product, which has become an important source of income. Automating internal research efforts has been described as the next frontier.
What a Researcher Will Actually Do at OpenAI
According to job listing Business ContentThe successful candidate can work in a variety of areas within OpenAI’s Prep team. These include defending OpenAI models against data poisoning, which involves attempts to corrupt an AI model through the dataset it is trained on, building tools to interpret the logic of models, and running experiments to understand the security implications of self-improving systems.
The researcher may also be asked to “track progress toward automation of technical staff,” including measuring how extensively AI coding tools are being used within the company.
The preparedness team’s broader remit includes preventing serious harm from AI. Other open roles in the same team; It includes automated red teaming to test for cyber vulnerabilities, biological and chemical risks, and threats posed by agency AI systems.
“This is an urgent, fast-paced effort with far-reaching implications for the company and society,” reads their Prep announcement, according to Business Insider.
Antropik Thinks Along Similar Lines
OpenAI is not alone in preparing for this change. In April, Anthropic published research on using AI models to audit more powerful AI models; The results were described as promising but limited. In May, Anthropic co-founder and chief policy officer Jack Clark wrote that he believed there was a roughly 60 percent chance of seeing AI research and development conducted without human involvement by the end of 2028.
“Any ‘reasonable’ civilization will take things much more slowly and carefully when it comes to artificial intelligence,” METR CEO Elizabeth Barnes wrote on Friday.
