I received many great questions from the community in response to my recent post on neuromorphic computing, so I’ll jump right in and answer a few. How does a… | 19 comments on LinkedIn
www.linkedin.com
View attachment 58928
I received many great questions from the community in response to my recent post on neuromorphic computing, so I’ll jump right in and answer a few.
How does a more powerful processor increase energy efficiency?
#AI is already used in advanced driving assistance systems (ADAS) and infotainment, and the complex calculations are currently performed on traditional CPUs, GPUs and NPUs, which are not energy efficient. #Neuromorphiccomputing requires for the same tasks less energy. As the number of AI functions continue to increase, the increased computing efficiency of neuromorphic hardware will require less energy in comparison to legacy hardware. Reduced energy usage will also increase vehicle range and improve sustainability.
When can I experience neuromorphic computing?
Widespread use of neuromorphic computing will depend on many factors. The technology requires new programming and algorithms, so it will not immediately replace traditional processors. One key factor for us is that automotive-grade chips must meet extremely strict reliability requirements. However, we are already actively working to drive development and we are committed to being the first to use this technology in the automotive industry.
If you haven’t read the article yet, check it out here
https://lnkd.in/epnUc5Sy. Be sure to ask more questions so we can keep the conversation going.
Neuromorphic computing? We’ve got that.
Because it’s still nascent technology, I am frequently asked to describe #neuromorphic computing. It is a paradigm shift for how we perform computations in machine learning (#ML) and artificial intelligence (AI), which process massive amounts of data requiring tons of fast memory.
Currently available processor architecture separates data calculations from system memory, which is inefficient. The biological inspiration for neural networks is the human brain, where computing and memory are combined, and data processing uses neurons to communicate through electrical signals and chemical processes known as neurotransmitters.
In neuromorphic computing, those human neurons and synapses are modelled in circuits and communication is event-driven, with information coded in spikes, mimicking the processing fundamentals of the brain. Those spikes propagate through a Spiking Neural Network of artificial neurons and synapses to predict results. Information processing is measured by spike rate or spike time instead of the number of calculations. Thus, neuromorphic chips are more energy efficient and have lower latency than conventional CPUs and GPUs. That means much faster computation using considerably less power.
However, this change in data processing also requires new software algorithms specifically designed to work with neuromorphic hardware. Existing algorithms can only partially leverage the many benefits of neural technology. Thanks to Valerij, Alexander, Christina in the Innovations & Future Technology area and the rest of our team for tackling this huge project!
𝗔𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀
Neuromorphic computing reduces the power required for advanced AI computation, which is useful in applications where energy is limited, like electric vehicles. However, we still need automotive-grade chips with neuromorphic technology before this technology becomes common in cars.
We at Mercedes-Benz AG are currently working on novel algorithms that take advantage of neuromorphic computing to improve the energy efficiency and performance of our cars. Our primary goals are to extend vehicle range, make safety systems react faster, and increase the number of #AI functions possible. In 2020, we already joined the #Intel Neuromorphic Research Community and since then we are continuously expanding our collaborations with other research partners and universities to ensure our software and hardware solutions continue to lead the industry.
It's an exciting time to be in the world of automotive technology. Please share any questions and comments below.