Wu Tsai Faculty Members attending an event at 100 College Street

Grand Challenges in Cognition

A coordinated research effort across the Wu Tsai Institute to address key fundamental challenges on the critical path to understanding cognition. Interdisciplinary teams of researchers work together to develop and execute moonshot projects on a medium time horizon.

Principles
Structure
Challenge 1
Timeline

Founding Principles

INTENTIONAL

Advance the Institute’s mission with purpose by pursuing collective scientific priorities

RADICAL

Embrace ambitious and risky ideas that could or would not otherwise be supported

TEAM-BASED

Create models for team science in academia infused with external partnerships

Challenge Structure

The Grand Challenges in Cognition initiative launches in Fall 2024 with an inaugural challenge, Foundation Models of the Brain (described below). We expect to launch an additional challenge within the next year and annually thereafter. Grand challenges reflect critical gaps in knowledge and methods identified in coordination with the Institute’s Faculty Members, Steering Committee, and External Advisory Board. Potential areas were introduced and refined at the Institute’s recent annual conference. Each challenge proceeds in two phases.

This phase extends across two semesters. The primary purpose of the first semester is social. Assembled teams get to know each other, learn about the state of play in the challenge area, and share disciplinary perspectives and expertise. These exchanges occur across a series of Institute-organized activities, including an introductory retreat, social dinners, and roundtables with outside experts elsewhere at Yale or in other sectors (e.g., industry, community, policy, law, etc.). Teams also have the opportunity to participate in a professional development workshop in team science.

The purpose of the second semester is to design a moonshot project. With administrative support from the Institute, teams organize working meetings and other activities to write a research proposal collaboratively. They should propose an ambitious yet actionable research plan with details on timeline, budget, and personnel. Awards are sized for rapid impact, which may include: direct funding; project management and staff support; in-kind access to Institute spaces and equipment (e.g., brain imaging, microscopy, compute clusters); Institute investments in infrastructure, software, and/or tool development; and fellowships for undergraduates, graduate students, and postdocs. Although the proposed research should be conducted at Yale, the involvement of outside individuals and groups is encouraged and supported. The Institute has existing and potential relationships with partners in several sectors that could augment resources, provide guidance, and increase the impact of research and engineering outcomes from these proposals.

For their time commitment and active engagement in Phase I, participants each receive $10,000 in discretionary research funds.

Research proposals submitted at the end of Phase I will be evaluated by a review panel and supported based on mission alignment and the availability of funding. Teams will execute accepted moonshot projects during Phase II, which we expect will last two years.

Each team project should yield standard metrics of scientific progress (e.g., papers, devices, and spinoffs) and will hopefully attract longer-term funding from federal agencies, foundations, or industry. However, the main measures of success will be scientific understanding, new collaborations, and serendipitous discovery. More details will be provided as teams reach this phase.

Challenge 1

Foundation Models of the Brain

Foundation Models of the Brain responds to the fact that neuroscience is awash in data but often short on insights about cognition. Data are collected with different tools, from different species or age groups, at different scales in space (molecules to entire brains) and time (milliseconds to years), and in different tasks. Worse still, data formats and standards vary across approaches, so each approach requires specialized knowledge and software. Few neuroscientists have expertise beyond a small number of data types; thus, large-scale integration of insights about cognition has proven difficult.

As used for AI, foundation models serve as a base of knowledge, aggregating diverse data sources to learn latent structures that can be extended creatively. Foundation models applied to the brain could be used to ingest many neuroscience data types: cellular and brain-wide imaging, oscillatory activity in neural populations, rich behaviors and task parameters, impacts of brain stimulation, pharmacology, or other perturbations on these behaviors, and even the text of scientific articles written about these datasets. The hope is that such integration will reveal general principles of cognition and help generate new interpretations and hypotheses.

Wu Tsai Faculty Members are invited to apply individually for Challenge 1.

APPLY BY JULY 31