The current era of artificial intelligence (AI) is presenting unique challenges for university researchers. At a recent convening for the Schmidt Sciences AI2050 program, I observed how academic AI experts are adapting to this new reality. They can study and observe AI technologies, but often lack access to the tools they need to conduct detailed research.

Professor Nika Haghtalab from UC Berkeley likened the situation to biologists in a world where gene-editing tools like CRISPR are controlled by private companies. Academics can examine the end results but cannot influence their development or design, which is crucial for progress in AI.

Funding remains a significant issue. While some fellows receive support from the AI2050 program to purchase GPUs, this relief is not enough to offset broader financial constraints on academic research. The cost of repeatedly querying models from companies like Anthropic and Google also poses a barrier.

Many researchers are focusing their efforts on areas that do not compete with corporate priorities, such as climate change or building specialized AI tools for specific applications. However, this shift has led to concerns about the broader impact of widespread misconceptions about what constitutes 'AI' in society. The ongoing threats and challenges are reshaping academic life, but there is also hope that these constraints could drive innovation.

Ultimately, while the landscape of academia may be changing, these researchers remain resilient and continue to seek ways to push the boundaries of AI despite the obstacles.