Accelerate Foundation Models Research (AFMR) is a new research initiative where we are on a journey, working together with the broader academic research community, to explore different aspects of foundation models to accomplish three goals:
Align AI with shared human goals, values, and preferences via research on models
which enhances safety, robustness, sustainability, responsibility, and transparency, while ensuring rapid progress can be measured via new evaluation methods
Improve human interactions via sociotechnical research
which increases trust, human ingenuity, creativity, and productivity, and decreases the digital divide while reducing the risks of developing AI which does not benefit individuals and society
Accelerate scientific discovery in
natural sciences
via proactive knowledge discovery, hypothesis generation, and multiscale multimodal data generation
Academic Research Community focus areas
Model Advancement
Projects aim to enhance LLMs’ alignment, robustness, interpretability, reasoning, and interaction abilities using various methods and techniques.
Benchmarks, Evaluation and Measurement
Develop metrics, evaluate and measurement of LLMs on various aspects of comprehension, benchmarking, and content generation, with expected outcomes of better usability, collaboration, robustness, and data science applications.
Multimodal and Crossmodal Learning
The research projects focus on improving and applying Multi-Modal foundation models.
Responsible AI
The projects aim to make AI more responsible by focusing on safety, preventing misinformation, and improving auditing in a way that’s easy to understand.
Cognition and Societal Benefits
Projects focus on advancing healthcare, education, legal and social aspects of using Large Language Models (LLMs).
Creativity and Design
How AI can be used across different disciplines to support human creativity, improve workflow efficiency, and enrich the user experience.
Multicultural Analysis and Empowerment
The research projects primarily focus on enhancing language models, emphasizing underrepresented languages and cultures.
Scientific Discovery and Innovation
How AI can be used to enhance scientific knowledge discovery and hypothesis generation across many different areas.
Domain applications
How foundation models can be applied in a variety for domain-specific applications in science and engineering, from agriculture to energy to health.