OpenAI's newest team is proof that there's a key shift happening in the AI talent wars

In a significant move within the AI sector, OpenAI has announced the formation of a new team called "Applied Evals," aimed at enhancing how businesses utilize artificial intelligence in complex processes. Led by engineer Shyamal Anadkat, this initiative will focus on applications such as handling refund requests and code migration, as well as advancing voice AI technologies and multi-step reasoning tasks. The recruiting post indicates that salaries for positions within this team range between $255,000 and $325,000, along with equity options. The concept of 'Evals' involves assessing AI model capabilities, which Anadkat highlighted as a crucial aspect of developing effective AI products. This strategic shift reflects a broader transformation in the AI talent landscape, where the emphasis is transitioning from purely technical expertise to including individuals with specialized knowledge in various fields. Business leaders in the AI sector have observed this trend, noting that while technical engineers have traditionally built and trained models, there is now a growing need for professionals who can apply these models effectively. Anadkat stated that the new team will initially consist of generalists, but plans are in place to incorporate specialists, such as software engineers for code-related tasks and potentially experts from the humanities for roles involving writing, depending on future demands. This team will collaborate closely with companies using OpenAI's developer platform, assisting them in creating tailored evaluations for their specific needs. "OpenAI began as a research lab, and we've transitioned into product development," Anadkat explained. "The goal now is to integrate research and product development more closely to define what quality means in practical applications." The focus of Applied Evals will be exclusively on business clients, operating independently from teams dedicated to consumer-facing applications. Anadkat mentioned that the team would prioritize projects based on customer needs and identify areas where AI models may be lacking. As the AI industry evolves, experts note that the approach to evaluating AI models has shifted from simple binary assessments to a more nuanced understanding of context and appropriate questioning. According to Michael Jacobides, a strategy and entrepreneurship professor, this shift is essential as AI moves towards more specific applications. Tanmai Gopal, CEO of PromptQL, emphasized that as AI becomes tailored to specific use cases, the definitions of success and failure become increasingly complex, highlighting the need for specialized knowledge in the evaluation process.