Genesis 2.0 Continuously Learning AI Chip from UT San Antonio
Researchers have conducted a five-year study to eliminate the issue of memory erasure experienced by current artificial intelligence systems.
Researchers at the Matrix AI Consortium within UT San Antonio have developed the Genesis 2.0 continuously learning AI chip, which solves the problem of artificial intelligences losing old information while learning new tasks.
Genesis 2.0 Chip and Development Process
Led by Dhireesha Kudithipudi, director of the UT San Antonio Matrix AI Consortium and Neuromorphic AI Laboratory, the research team worked to prevent current AIs from overwriting previously learned information when programmed with new tasks.
Architecture Inspired by the Human Brain
The Genesis 2.0 AI chip, the product of a five-year study, incorporates continuously learning software inspired by the way the human brain forms connections and was designed accordingly.
Technical Infrastructure and Team Contributions
Team members such as Nicholas Soures, Fatima Tuz Zohora, and Vedant Karia brought capabilities to the project including spiking neural networks, energy-efficient nanostructured memory memristors, and low-power accelerators.
Funding and Future Goals
Costing approximately $2 million, the project was funded by a grant provided by the Air Force Research Laboratory, and the chip continues to be developed before becoming ready for real-world applications.