AI Helps Students Build Technology Skills

Summary:

The University of Florida (UF) adopts an approach offering a roadmap for the seamless integration of AI across various disciplines, improves academic quality, and strengthens UF’s reputation for research and innovation.

Client: 

The University of Florida (UF) is a public institution located in Gainesville, Florida. It comprises 16 academic colleges that offer a diverse array of arts and sciences programs.

Problem Statement: 

Only 60% of employers believe that new graduates have the knowledge and skills needed for success in entry-level tech positions. The global tech talent shortage is expected to reach 85 million people by 2030, leaving many roles unfilled and research projects understaffed. In response, the University of Florida (UF) has partnered with NVIDIA to leverage AI hardware and expertise, adopting a comprehensive approach to AI education. The university aims to conduct pioneering research to attract top talent, students, and investments.

UF’s AI steering committee has invited all 16 colleges to collaboratively plan and fund AI initiatives, aiming to offer equal opportunities for students to prepare for and engage in an AI-driven future.

Results: 

• 100 AI-focused faculty members.
• The university offers a nine-credit AI certificate for undergraduates to acquire a baseline of AI literacy. 
• Overcoming internal resistance by distributing costs and benefits across the institution.
• Engaging with local government and private donors to secure financial support.
• Launching ambitious research projects to attract top talent and motivated students.
• Extending AI knowledge and tools beyond the campus through partnerships and open-enrollment programs.

AI Solution Overview:

NVIDIA installed HiPerGator AI, an NVIDIA DGX SuperPOD™, at the University of Florida (UF). This setup includes 140 NVIDIA DGX™ A100 systems, each with eight NVIDIA A100 Tensor Core GPUs and an additional 17,920 CPUs. Completed in just 90 days, it became the largest AI supercomputer ever installed at a university.

Using HiPerGator AI, UF developed GatorTron, the largest clinical language model to date, featuring 3.9 billion parameters. GatorTron is over 10 times larger than the second-largest clinical model and reduced misclassification by 24% compared to BERT-large in the i2b2-2010 named-entity recognition benchmark.

Since its deployment in January 2021, HiPerGator AI has been instrumental in advancing AI education, conducting impactful research, and preparing students for in-demand jobs. To foster AI engagement across various disciplines, UF established the AI Research Catalyst Fund. The first funding round received 133 proposals, from which 20 were selected for their potential to enhance the university’s research profile. Funded projects include:

– AI for detecting biomarkers for Alzheimer’s disease

– Machine learning for tracking land use and development impacts in Florida

– AI for identifying academically at-risk students

– AI for detecting parasitic nematodes affecting agriculture

UF combined its existing data analysis strengths with new AI computing capabilities to drive these projects forward. HiPerGator AI was also used to train the GatorTron NLP model on 82 billion words and 8.9 million parameters from medical records in just seven days, surpassing the previous largest model, ClinicalBERT.

In a follow-up project, UF and NVIDIA are developing SynGator, a neural network designed to generate synthetic clinical data. This tool will create digital health records for training healthcare AI models, helping researchers develop tools without compromising protected health information.

References: 

1. Bringing AI Education to Every College and Discipline. https://www.nvidia.com/en-us/case-studies/bringing-ai-education-to-every-college-and-discipline/

Industry: Education Services

Vendor: NVIDIA

Client: University of Florida

Publication Date: 2021