Backpack to the Future

Tiny machine learning, or tinyML, has emerged as a disruptive technology that is transforming the landscape of artificial intelligence and edge computing. This innovative approach involves deploying machine learning models onto resource-constrained and low-power devices like microcontrollers. The rise of tinyML can be attributed to several factors, including the increasing demand for localized and real-time AI applications, the proliferation of small, low-power devices, and advancements in model optimization techniques.

One of the key reasons tinyML is gaining prominence is its ability to bring intelligence to a wide range of devices that were previously unable to support such capabilities. Smartwatches, smartphones, fitness trackers, and even IoT devices are now capable of running ML applications. This democratizes AI and makes it accessible to a broader audience while reducing latency and enhancing privacy by processing data locally rather than relying on cloud-based solutions.

The technology has led to the development of many successful applications in areas such as health monitoring, predictive maintenance, and environmental sensing. However, it is important to acknowledge the limitations of tinyML on highly resource-constrained devices like smartwatches. While the applications are incredibly useful and continually improving, there are inherent constraints, such as limited memory, processing power, and energy efficiency. Complex models and computationally intensive tasks may still pose challenges on such devices, requiring a delicate balance between accuracy and efficiency. Additionally, training and deploying models must consider these limitations and involve careful model compression and optimization to ensure effective performance on these devices. And for more resource-intensive applications, traditional portable and wearable devices are not yet up to the challenge.

Based upon an application recently accepted by the United States Patent and Trademark Office, Microsoft might be up to something that enables more powerful AI applications to be used on the go. To get some more breathing room in terms of size, weight, and energy consumption constraints, Microsoft proposed the development of a smart backpack. Such a device could pack a lot more horsepower than a smartwatch or smartphone without being too cumbersome for daily use.

To be clear, Microsoft has not announced if an AI-powered backpack is actually under development at this time, so if and when such a device comes to be, it is not certain exactly what it might be capable of or what hardware it will contain. But the application hints that Microsoft is looking at installing forward-facing sensors, like a camera, microphone, GPS receiver, and compass in the backpack’s straps. There may also be speakers, haptic actuators, and LEDs incorporated into the straps to provide the wearer with a variety of types of feedback.

Mentions of onboard processing are also made, so it is probably safe to assume there will be computing resources, and enough battery power to keep them chugging along. And since we assume they chose the backpack form factor for a good reason, it should be safe to say that the computational resources will be pretty substantial. While it should be able to handle most workloads without relying on the cloud, wireless connectivity is also mentioned, which could support a number of use cases.

It would appear that Microsoft envisions this backpack as being a sort of personal AI assistant. Towards that goal, it will support image, text, speech, facial, and cognitive recognition capabilities. This could provide the wearer with contextual insights, advanced object recognition, and an ability to interact with nearby devices.

Among the applications the backpack might ultimately support are some normal daily activities, like navigation, buying concert tickets, or checking prices at the grocery store. This form factor offers a lot of potential for AI applications, but only time will tell if this AI-powered backpack will ever emerge.