CHAT TO US ABOUT:
- Reduction of power for edge & cloud temporal AI
- Our developer studio, Nengo
- How we can help you achieve your SWaP-C goals!
MEET OUR TEAM:
Peter Blouw
Senior Research Scientist
Chris Eliasmith
CO-CEO
Peter Suma
Co-CEO
ABOUT US:
Description
We are AI engineers and neuroscientists who have taken algorithms found in the brain and applied them to AI computing problems to make more efficient AI for edge, cloud and autonomous applications.
Our patented temporal AI model called a Legendre Memory Network (LMU) is that much better than the current LSTM, CNN or other AI models in use for temporal AI processing.
For cloud computing uses, a simple software change to replace LSTM's & CNN models with our LMU models is all it takes to get a 40% reduction with no hardware change, and often a 10x increase in speed. This means less hardware, less energy, less compute time.
For edge users such as makers of smart-phones, smart-speakers, ear-buds, smart-glasses, ioT sensors and wearable computing devices we have software and also ready to go embedded chip designs. For ear pieces, and clothing where battery power is super scarce, we can reduce your battery usage for automatic speech processing (ASR) by 16x and speed up response by up to 10x! Imagine what you can do with all the extra AI processing time and power left over. You can make your devices the smartest in the market compared to those wasting energy on inferior solutions.
Examples of Applications Built with Our Technologies:
1. Edge Speech: Our speech recognition technology for example uses 94% less power, running on 9 micro-watts, compared to all other solutions which start at 140 micro-watts. This means you the batteries in your devices last much longer and you can compute larger, more useful AI networks, making your devices smart than the competition's, all for a few cents per device. Using our low-power AI technologies, we also build cognitive autonomous 'brains' for devices, like fully-autonomous drone controllers for tasks such as infrastructure inspection and military control R&D.
2. Autonomous Drones: Our autonomous drone control system uses our low-power computing algorithms to deliver combined adaptive flight control, path planning and visual detection all computed onboard to deliver integrated intelligent behaviors for infrastructure inspection. It doesn't get much more demanding than the power, response and intelligence demands of this kind of fully integrated, real-time solution. ABR is a leader in autonomy, coming from brain modeling and has had to build the world's best algorithms for temporal sequence learning and intelligent processing.
Our patented temporal AI model called a Legendre Memory Network (LMU) is that much better than the current LSTM, CNN or other AI models in use for temporal AI processing.
For cloud computing uses, a simple software change to replace LSTM's & CNN models with our LMU models is all it takes to get a 40% reduction with no hardware change, and often a 10x increase in speed. This means less hardware, less energy, less compute time.
For edge users such as makers of smart-phones, smart-speakers, ear-buds, smart-glasses, ioT sensors and wearable computing devices we have software and also ready to go embedded chip designs. For ear pieces, and clothing where battery power is super scarce, we can reduce your battery usage for automatic speech processing (ASR) by 16x and speed up response by up to 10x! Imagine what you can do with all the extra AI processing time and power left over. You can make your devices the smartest in the market compared to those wasting energy on inferior solutions.
Examples of Applications Built with Our Technologies:
1. Edge Speech: Our speech recognition technology for example uses 94% less power, running on 9 micro-watts, compared to all other solutions which start at 140 micro-watts. This means you the batteries in your devices last much longer and you can compute larger, more useful AI networks, making your devices smart than the competition's, all for a few cents per device. Using our low-power AI technologies, we also build cognitive autonomous 'brains' for devices, like fully-autonomous drone controllers for tasks such as infrastructure inspection and military control R&D.
2. Autonomous Drones: Our autonomous drone control system uses our low-power computing algorithms to deliver combined adaptive flight control, path planning and visual detection all computed onboard to deliver integrated intelligent behaviors for infrastructure inspection. It doesn't get much more demanding than the power, response and intelligence demands of this kind of fully integrated, real-time solution. ABR is a leader in autonomy, coming from brain modeling and has had to build the world's best algorithms for temporal sequence learning and intelligent processing.