TDK Corporation, a leading electronics company, has made a significant breakthrough in AI technology by successfully collaborating with the French research institute CEA and Tohoku University in Japan. The collaboration commenced in 2020, focusing on developing an innovative memristor technology, and has now led to the creation of advanced memristors integrated into an Integrated Circuit (IC).
These memristors were produced utilising the semiconductor processes developed at the Center for Innovative Integrated Electronic Systems (CIES) at Tohoku University. Notably, Tohoku University boasts a sophisticated 300mm-compatible spintronics semiconductor prototyping line, which was pivotal in this process.
The newly developed integrated memristors exhibit exceptional characteristics: they can be controlled easily, showcase minimal current leakage, and demonstrate enhanced resilience against resistance drift over time, which is crucial for long-term data storage. TDK has also attested to their immunity to environmental impacts, further underpinning their robustness.
In a noteworthy collaboration, TDK and CEA have built an AI circuit utilising four of these innovative chips. This circuit is powered by an analogue algorithm specifically designed to learn and differentiate between three types of sound - music, speech, and noise. The technology's standout feature is its ability to perform real-time learning in dynamic environments, a departure from traditional machine learning models that rely on pre-trained data sets. Remarkably, the system demonstrated an ability to successfully distinguish and learn to separate sound types even when mixed in varying proportions.
One of the key advantages of using analogue algorithms, such as the one developed here, is their substantially lower power consumption compared to conventional digital neural network implementations. TDK reports that the power consumption is as much as 100 times lower. This capability is emphasised by drawing parallels to the human brain, which operates on approximately 20 watts yet is capable of making more complex decisions than current digital AI processors with significantly less energy.
Following the successful demonstration of their AI circuit, TDK, in conjunction with Tohoku University, plans to explore integrating semiconductor and spintronic manufacturing processes. This integration has been successfully achieved in the production of Magnetic Random Access Memory (MRAM) - a product similar to memristors - with Tohoku University at the forefront of MRAM research and development.
The advancements made in this project have the potential to significantly impact the future of AI technology and memory storage solutions, presenting opportunities for more efficient and adaptable AI systems.
Source: Noah Wire Services