Many businesses are turning to artificial intelligence to improve efficiency and productivity. Tiny Machine Learning changes the way we gather and use data, enabling real-time data capture.
Artificial intelligence is taking the world by storm. Through its recent advancements, businesses are learning how to use AI to be more efficient. Efficiency breeds higher productivity, which breeds higher profit margins. A very key piece of AI is Machine Learning, it’s the part of AI that learns. Typically, ML is trained on sample data, information gathered from around the web. Now, a new version of ML is here. TinyML changes the game by allowing machines to learn and grow through use.
Machine learning algorithms for hardware require complex mathematical modes based on training data. It’s not only complex, but expensive to build. Not to mention that tasks related to ML are typically translated to the cloud, which creates latency and consumes power and compute. This strains the system and forces the machines to rely on connection speeds, making the process slower and less predictable.
TinyML is embedded software technology that enables machines to learn and grow through use, there’s no pre-loaded training data, the code is driven by the data you are feeding it in real time. It’s the latest trend in building product intelligence, and the product of a collaboration between Arduino and Edge Impulse, which got a kickstart from Kartik Thakore, who provided the dataset and started the project in an effort to help with COVID-19 efforts. Now, TinyML is available on open-source repository, Hackster.io.
The partnership between Arduino and Edge Impulse allows developers to run powerful learning models based on artificial neural networks. The ability to reach and sample tiny sensors along with low-powered microcontrollers is the driving force behind this technology. Huge strides have been made in deep learning models, and TinyML is the bridge between Edge hardware and device intelligence. The result is tiny devices with not-so-tiny brains.
The implications of this technology are vast. It can be used to shorten drug development from about five years to just 12 months. It can be used to detect anomalies in insects as small as wasps, which could trigger a sensor to alert farmers so they can provide assistance, in real-time. Edge Impulse and Arduino already published a project illustrating how TinyML running on an Arduino Nano BLE Sense can detect specific coughing sounds. It also includes a dataset of coughing and background noise samples, everything in real-time. The cough detection system runs in under 20KB of RAM on the Nano BLE Sense.
The best part about TinyML is that it runs on small devices without a need to send data to the cloud. Previous ML models are huge, far too big to fit onto a mobile device. But TinyML is small enough to fit into any environment, it runs locally and performs on-device analytics, which is part of the reason it’s becoming so popular. Accessibility is another reason TinyML is becoming popular, software developers and engineers can now build embedded systems using ML by making their mobile device the edge device. Being able to tap that device anywhere is huge, especially in the remote work era.
Companies have a need for easily deployable and inexpensive solutions, especially amid the pandemic. TinyML has implications well beyond what is mentioned here, most of which will be unknown until someone pioneers that path.
This is a technology for businesses to watch. It may not be practical for every project or every situation, there will be times when you need the full power of traditional ML, but TinyML has the potential to change the way we use technology in the future. As you watch the developments using TinyML, be on the lookout for an expert in ML. If you decide to go that route, you’ll want to have one on speed dial to make sure it’s implemented and used appropriately.