Tweet
The potential impact that artificial intelligence (AI) will have on society as a whole cannot be overstated. The challenge of widespread AI still persists, though. Processing power, time, and resources can all be used up quickly during algorithm training. Robotics and medical imaging applications will raise the burden of inference. AI could consume up to 25% of the world's electricity, according to Applied Materials, if we don't make improvements in processors, software, material science, system design, networking, and other areas. The two main strategies for computing and AI technology development nowadays are extreme-scale systems and edge/pervasive massively distributed systems. Both have a combination of similar and different problems.
Here are the key areas where I believe hardware needs to be enhanced.?
1. Specialized Processing
2. Near Data Processing
3. Non-CMOS Processors
4. Event-Based/Threshold Processing
5. Neuromorphic Processors
6. Extreme Ambient Cooling
7. Zero Compute Architectures
?