AI Companies Offering 500% Salary Hikes to PhDs Spark Academic Brain Drain

AI companies are increasingly offering substantial financial incentives to newly-minted PhDs, sparking concerns about an academic 'brain drain'. Larry Birnbaum, a professor of computer science at a prominent university, highlighted the challenge of competing with tech giants like Google, which can offer salaries up to five times that of a professor. This trend has intensified as industry salaries have surged, with reports of Meta offering seven- to nine-figure salaries to highly-experienced AI researchers, which in turn is pulling up the salary levels of even newly-minted PhDs.

Some academics fear that this trend is depleting the ranks of academic AI departments, which are crucial for important research and training the next generation of PhD students. Anasse Bari, a professor of computer science and director of the predictive analytics and AI research lab at a major university, expressed concern that the corporate opportunities available to AI-focused academics are significantly affecting academia. He emphasized the importance of investing in a solid AI education that upholds ethical values and cultivates thoughtful AI practitioners.

Before the rise of ChatGPT, top AI researchers were already in high demand. Many corporate AI labs, such as OpenAI, Google DeepMind, and Meta’s FAIR, allowed established academics to keep their university appointments, at least part-time. This model has reportedly declined due to intense talent competition, with companies offering millions of dollars for full-time commitment, which outpaces university resources and shifts focus to proprietary innovation.

While some professors argue that academia is a thriving component of this booming labor market, others are concerned about the impact on academic research. Henry Hoffman, who chairs the Department of Computer Science at a leading university, has watched his PhD students get courted by tech companies since he began his professorship. He mentioned a star student who dropped out of the PhD program to accept a high six-figure offer from a tech company.

The job market for computer science and engineering PhDs who study AI is in stark contrast to the one faced by undergraduate computer engineers. Many of those with bachelor degrees in computer science traditionally find jobs as coders, but large language models are now writing large portions of code at many companies. Meanwhile, most AI-relevant PhD students have their pick of jobs in academia, tech, and finance. These graduates are courted by the private sector because their training propels AI and machine learning applications, which can increase revenue opportunities for model makers.

AI and machine learning are the most popular disciplines among engineering PhDs. The all-but-assured path to prosperity has made relevant PhD programs in computer science and math extremely popular. Applications to AI PhD programs are on the rise, with some universities seeing a surge in admissions applications. This trend is likely to continue as the demand for AI expertise grows.

Despite federal funding cuts to universities, AI-related research is less affected because some of this research is funded by corporations. For example, Google is collaborating with a university to research trustworthy AI. This dichotomy underscores the decision of a major university to open its Data Science and AI Institute, a significant effort to enroll PhD students in engineering disciplines and hire new tenure-track faculty members.

For some academics, the reasons to remain in academia are ethical. Luís Amaral, a computer science professor, is concerned that AI companies have overhyped the capabilities of their large language models and that their strategies will breed catastrophic societal implications. He believes that academic labs are among the few places actively exploring alternative AI architectures beyond large language models and transformers, making academia's role as a hub for non-mainstream experimentation more important in this corporate-dominated landscape.