Artificial intelligence (AI) has been a rapidly growing field, with researchers constantly pushing the boundaries of what is possible. One area that has seen significant advancements is the development of neuromorphic chips, which aim to mimic the functioning of the human brain. And now, a team of electrical and computer engineers at Johns Hopkins University has made a groundbreaking discovery in this field.
The team, led by Professor Andreas Andreou, has successfully designed a neural network chip using natural language prompts and ChatGPT4. This marks the first time that an AI chip has been designed by a machine using natural language processing. The process is similar to telling a computer to make an AI neural network chip and having it generate a file that can be used for manufacturing.
The work was initiated at the 2023 Neuromorphic Cognition Engineering Workshop and has recently been posted on the preprint site arXiv. The final design of the chip consists of two layers of interconnected neurons, with an 8-bit addressable weight system that allows for user-adjustable connections. The team also developed a user-friendly interface called the Standard Peripheral Interface (SPI) sub-system, which acts as a remote control for reconfiguration and programmability.
Michael Tomlinson, an electrical and computer engineering graduate student and member of the Andreou Lab, explained that they intentionally kept the design simple as a proof of concept. Before sending it for manufacturing at Skywater foundry, they conducted extensive software simulations to ensure its functionality.
This breakthrough not only showcases how AI can be utilized in creating advanced hardware systems but also paves the way for future developments in this field. As Tomlinson puts it, while this is just a small step towards large-scale automatically synthesized practical hardware AI systems, it demonstrates that AI can be employed to create advanced AI hardware systems that in turn would help accelerate AI technology development and deployment.
The semiconductor industry has made significant progress in scaling down the size of chips over the years, but this discovery takes it a step further by utilizing AI to design these chips. This not only saves time and resources but also opens up new possibilities for the future of AI technology.
With this breakthrough, we can expect to see more advancements in neuromorphic chips and their potential applications in real-time machine intelligence for embodied systems such as autonomous vehicles and robots.
Source: techxplore.com
The ENTECH magazine, aimed at teenagers interested in STEM education, will also feature this groundbreaking discovery, inspiring young minds to explore the world of AI and its endless possibilities.
To stay updated on this exciting development and other STEM-related news, visit ENTECH magazine today!
