New AI Trick That Could Make Your Apps Smarter and Faster
Your phone’s AI might start working better without needing more power...
Researchers just unveiled a new method for training AI that avoids the usual heavy math. Think of it like a sharper ruler for measuring things: instead of judging by fixed steps (additive margin), this method compares things by a fraction of their size (multiplicative margin), making it more stable no matter the data.
The new approach, called MMPerc, works like a smarter version of the old ‘Perceptron’—a simple AI brain used in everything from spam filters to facial recognition. It’s faster to train, needs less computer power, and works well even on small devices like your phone or smart watch. The team tested it on real-world data and found it outperformed older AI methods.
What’s really cool is that this could help AI run better on devices you use every day. Imagine your phone’s AI assistant understanding your voice commands faster, or your smart thermostat learning your habits quicker—all without needing to send data to the cloud. That means less waiting and fewer privacy worries.
The researchers say this method is especially promising for resource-strapped situations, like in medical devices or battery-powered gadgets. It’s like giving AI a diet plan: simpler, lighter, but still effective.
- New AI training trick (MMPerc) is faster and uses less power than older methods like support vector machines
- Works better on small devices like phones or smartwatches, not just giant computers
- Could improve everyday AI tools like voice assistants, spam filters, and face recognition
Why It Matters
AI that works better on your phone or watch, using less power and keeping your data local