Memristor Kit For Artificial Intelligence Chips: Nöral Timing Applies!
Automatically translated
CreatorArtificial intelligence (AI)The most exciting and potential of the advanced technology of our era is one of the highest areas. In this field, a study conducted in the leadership of the University of Michigan, is a great breakthrough for AI chips. In this study, researchers are able to process changing data in time much more efficiently and thus increasing the overall data processing capacity of AI significantly memristor they developed. Homeis an electronic component that stores electrical resistance and can change this resistance.
The name comes from the combination of words “memory” (bellek) and “resistor”. Memristors can greatly improve the performance and energy efficiency of artificial intelligence chips in the future. This technological development can greatly reduce the energy consumption of AI and this can lead to the use of AI systems in a wider and effective way. This invention can be an important step in increasing the sustainability and efficiency of AI technology.
Neurons and Memristors: Timing Mechanism
The neurons in our brain follow the time to relax in different speeds after the signals they receive. This allows them to better understand the world and time-changing events in our surroundings. Now, memristors - hardware analogues of neurons - have also made the same thing. This means that artificial neural networks can process changing information over time, such as audio and video, more efficiently. A team led Wei Lu and John Heron from the University of Michigan, developed and reported the first memristorue, which can be adjusted from the 'decoration time'. This is considered a great progress in science and technology. This new technology can ensure computers and artificial intelligence not only data, but also the flow of time better.
You can withdraw your interest: A Artificial Intelligence Hospital Town was opened for the first time! 10,000 Patients Treat it in a Week!
Energy Saving and the Future of AI
Memristors are revolutionary electrical components that store information in electrical resistance. These components can reduce the energy requirement of AI in serious sense, so it can reduce approximately 90 times according to today's graphic processing units (GPU). This is extremely important both environmental and financially. Energy consumption of these technologies is a concern that grows rapidly. So, when it comes to 2027, it is estimated to form half a percentage of world electricity consumption. This rate can potentially increase with more companies’ AI tools to develop, sell and start using. Therefore, increasing energy efficiency is vital for the sustainability of our planet.
New Material System and Training Process
Although Wei Lu’s team had investigated adding relaxation time to memristors in the past, this was not a systematically controlled phenomenon. This situation was difficult to provide a certain layout and control over the functionality of memristors. But now, the team of Lu and Heron revealed memristor networks that can mimic this timing mechanism. These networks provide variations on a basic material that may have different relaxation times, offering a level of control that can not be obtained before. So, the memristors have a certain relaxation time, and this time can be controlled, it is seen as a big step for new technological developments.
Meets Revolution in Artificial Intelligence Chips
These memristors were manufactured with an intensive process that requires excellent crystals, in order to measure the characteristics of the team precisely. This process requires careful control of each detail. However, the team predicts that in order to make this technology suitable for mass production, a much simpler and efficient process will work.
Artificial Intelligence Aided Brain Implants with Neuralink: A New Hope Light for Patients with Felçli
This process can eliminate the need for creating perfect crystals and this can speed up the production process. The team member Heron calls such oxides as the "mutfak basin" of the atom world. This means oxides become more stable with increasing the number of added elements. The team achieved time constants ranging from 159 to 278 nanoseconds, one of the trilyonda of seconds by changing different oxide rates. This is a critical step to improve the performance and availability of memristors.
Future Vision
By installing a simple memristor network, they learned to recognize the sounds of numbers up to nine from scratch. After grafted, voice input has been recognized every number without completion. This progress can expand the applications in different areas of AI, with the potential to increase the energy efficiency of artificial intelligence chips. This technology can play an important role in the development of AI, enabling future AI systems to work faster and efficiently.
Creator
