AI in the developing world: how ‘tiny machine learning’ can have a big impact
The landscape of artificial intelligence (AI) applications has traditionally been dominated by resource-intensive servers centralized in industrialized nations. However, recent years have witnessed the emergence of small, energy-efficient devices for AI applications, known as tiny machine learning (TinyML).
TinyML devices, with their small size and low cost, have the potential to make a big impact in the developing world. These devices can be deployed in the field for various applications. For example, they have been used to detect mosquito wingbeats, aiding in the prevention of malaria spread. They have also played a role in the development of low-power animal collars to support conservation efforts.
What makes TinyML devices unique is their small size and affordability. Unlike traditional AI systems that require servers or smartphones, TinyML devices can run on simple microcontrollers that power standard electronic components. With 250 billion microcontrollers already in use globally, devices that support TinyML are readily available at scale.
There are several development packages available for TinyML applications, including Arduino and Seeed Studio. These packages come with additional sensors for audio, vision, and motion-based applications.
So how does TinyML work? Like classical machine learning, TinyML involves data collection, often from Internet of Things (IoT) devices, and cloud-based training. However, in the TinyML system, the trained model is deployed on the device itself, enabling real-time data analysis and decision-making without the need for constant connectivity to the cloud.
TinyML offers several advantages over traditional centralised server-based models. Firstly, its low cost makes these devices accessible to a wide range of users, including educational institutions and students in the developing world. Secondly, the modest energy consumption of TinyML devices reduces their impact on the environment, making them a sustainable choice. Additionally, TinyML enables the development of applications that specifically address the needs of local communities, while also providing the ability to operate without internet connectivity, which is particularly beneficial for regions with limited access to the internet.
TinyML applications have already been used in personalized sensors for athletics, providing localization where GPS is not available, and by startups to offer privacy-conserving conversational agents and person-detection hardware. The low-cost, low-power microcontrollers make these applications possible.
To promote the use of TinyML in the developing world, a network called TinyML4D has been established. This network includes academic institutions from over 40 countries in the Global South, aiming to improve access to TinyML education and develop innovative solutions to address the unique challenges faced by developing countries. Efforts have been made to distribute TinyML hardware kits to universities with budgetary challenges and organize workshops and training sessions. The network also focuses on collaborating with practitioners and addressing the United Nations’ sustainable development goals (SDGs).
While the growth of TinyML devices and applications presents exciting opportunities, there are also potential challenges and risks. The increasing number of devices could lead to electronic waste due to their low-cost nature. Additionally, embedded biases in ML models and concerns about privacy need to be addressed responsibly.
Overall, TinyML represents a transformative approach to AI, particularly in the developing world. It offers a sustainable path to democratizing AI technology, fostering local innovation, and addressing regional challenges. As the field evolves, it will be crucial to navigate the potential drawbacks and ensure that TinyML remains a tool for positive change and sustainable development.
Source: The Conversation