Explainer
TinyML brings small machine-learning models to devices with tight memory and power limits. A local model can detect a pattern, such as a wake word, without sending every sample to a distant server. The device can then act on that result or send selected information onward when the application needs it.
The key ideas
- Small models can run on constrained hardware.
- Local inference can reduce traffic and delay.
- A local detection task can coexist with cloud processing.
Explore the subject
Original recording: TikTok, 2021-01-25.
Video transcript
Read the original recording’s transcript
Automatically transcribed from the original recording. Small transcription errors may remain.
What is TinyML and what is so special about it? Well, TinyML is a method and field of learning that puts artificial intelligence in your home or your pocket without the hurdle of sending large chunks of data to the cloud, helping avoid latency issues and slowing down the internet, which significantly improves user experience. So this is done by embedding machine learning algorithms on small, low-powered hardware devices like your smartphone, your watch, or your Google Assistant. In exchange, much smaller chunks of data are sent to the cloud if needed for further processing since most of the work is done locally
on the hardware itself. An example of this is when you use wake words like Alexa to have your Amazon Echo respond instantly rather than waiting on internet infrastructure constraints. So for more context on this video, make sure you watch my other videos on cloud computing and the internet of things.
