The mission of the Kempner Institute for the Study of Natural and Artificial Intelligence at Harvard University is to understand the basis of intelligence in natural and artificial systems. The Institute will strive collectively to reveal the fundamental mechanisms of intelligence, endow artificial intelligence (AI) with features of natural intelligence, and apply these new AI technologies for the benefit of humanity.
The Institute will recruit and train future generations of researchers to study intelligence from biological, cognitive, engineering, and computational perspectives. It will elevate Harvard University’s standing in AI to the top ranks while expanding the national capacity to nurture new talent and develop expertise in a critical area of science and technology. Thus, the Institute will:
- Advance a fundamental and unified understanding of the nature of intelligence that has evolved in organisms and can be engineered in silico
- Train the next generation of researchers to work collaboratively across disciplines and to become versed in relevant areas of computational theory, AI and machine learning (ML), neuroscience, and cognitive science
- Translate theoretical advances to the development of next generation AI systems with qualitatively better capabilities for deeper reasoning and more efficient use of data
- Use knowledge gained from the study of AI systems to explain how the brain processes information, computes, learns and guides action
- Contextualize research via application challenges from complex domains of importance to humanity
Kempner Institute members are expected to model and support the following:
- Rigor in thought and research
- Constructive collaboration and community building
- Inclusivity, respect, and honesty in communication and action
- Curiosity to discover, connect, and understand
- A commitment to make the world a better place
The Kempner Institute is building a state-of-the-art computer cluster to support advanced research on intelligence from biological, cognitive, and computational perspectives.
The Kempner’s computer cluster will incorporate a range of hardware and software technologies— including specialized machine learning accelerators and advanced networking capabilities— to enable large-scale experimentation among researchers. Using this computing power, the Kempner’s scientists will be able to train state-of-the-art machine learning systems and analyze large data sets, both of which are critical to advancing research on the study of intelligence.
The Kempner will build its computing capacity over time. In the spring of 2022 the Kempner opened its computing cluster with an initial pilot installation of 144 NVIDIA A100 accelerators. Once fully assembled, the cluster will consist of the equivalent of approximately 800-1000 A100s or the contemporary equivalent. Once the full cluster is operational, the Kempner plans on annual upgrades and retirement of older machines to ensure that it stays up to date and maintains the compute capacity to support continuous cutting-edge research.
The Kempner’s initial cluster of 144 A100-40GB GPUs from Lenovo consists of 32 liquid cooled 4-GPU nodes that contain the necessary central processing units (CPUs), memory, and networking to perform state-of-the-art deep learning up to the level of the entire cluster. It can train a GPT-3 quality model in 55 days. The nodes will reside in the Massachusetts Green High Performance Computing Center (MGHPCC), a modern datacenter in western Massachusetts shared by the Boston-area universities.
In addition to building the computer cluster itself, the Kempner is also assembling a team of software engineers who will help researchers and students make the most effective possible use of the cluster and share their work open-source.
The Institute was launched in December of 2021, funded by a $500 million gift from the Chan Zuckerberg Initiative foundation. The institute’s name honors Mark Zuckerberg’s mother, Karen Kempner Zuckerberg, and her parents — Zuckerberg’s grandparents — Sidney and Gertrude Kempner.