Scholars & Big Models: How Can Academics Adapt?

In the wake of big vision models' success, computer vision has experienced rapid growth and increased attention, raising concerns about its impact on the CVPR academic community. Grad students feel discouraged because they lack access to immense compute in academia that power these big models, while senior researchers witness exponential growth beyond their wildest dreams. We now find ourselves at a pivotal moment in the history of CVPR, prompting several critical questions:

• How can we discuss and address concerns arising from the rapid changes due to the massively distributed training of big models?

• As big models become more prevalent, what can be done to reduce the barriers for entry, open access, and equity of opportunities?

• When several SOTA are being pushed every two weeks, how can the community support PhD students and researchers not feel discouraged?

• Are a few research directions affecting the questions the majority of our community focuses on?

• What creative solutions can benefit graduate students, assistant professors, senior faculty members, and other community members?

• How can we redesign collaborations, access to computational resources, and curricula to adapt to this evolving landscape?

This workshop brings together researchers from academia and industry, featuring diverse talks and panels to address these questions. Our speakers and panelists will share candid thoughts on the challenges and opportunities posed by big models. To improve inclusion, equity, and diversity in the future, we hope to gain a deeper understanding of the current landscape and possible strategies to adapt to and thrive in it.

Thanks to Derek Hoiem and Svetlana Lazebnik for expressing interest and sharing feedback in early brainstorming. We were encouraged by previous initiatives at CVPR on affecting our research community positively: Good Citizen of CVPR and Computer Vision After 5 Years.