The last decade has seen a surge in commercial applications using machine learning (ML). Similarly, marked improvements in latency and bandwidth of wireless communication have led to the rapid adoption of cloud-connected devices, which gained the moniker Internet of Things (IoT). With such technology, it became possible to add intelligence to sensor systems and devices, enabling new technologies such as Amazon Echo, Google Nest, and other so-called "smart devices." However, these devices offer only the illusion of intelligence and are merely vessels for submitting and receiving queries from a centralized cloud infrastructure. This cloud processing leads to concerns about where user data is being stored, what other services it might be used for, and who has access to it.7 More recently, efforts have progressed in dovetailing the domains of IoT and machine learning to embed intelligence directly on the device, known as tiny machine learning (TinyML).10

Machine learning (ML) sensors offer a new paradigm for sensing that enables intelligence at the edge while empowering end-users with greater control of their data. As these ML sensors play a crucial role in the development of intelligent devices, clear documentation of their specifications, functionalities, and limitations is pivotal. This paper introduces a standard datasheet template for ML sensors and discusses its essential components including: the system's hardware, ML model and dataset attributes, end-to-end performance metrics, and environmental impact. We provide an example datasheet for our own ML sensor and discuss each section in detail. We highlight how these datasheets can facilitate better understanding and utilization of sensor data in ML applications, and we provide objective measures upon which system performance can be evaluated and compared. Together, ML sensors and their datasheets provide greater privacy, security, transparency, explainability, auditability, and user-friendliness for ML-enabled embedded systems. We conclude by emphasizing the need for standardization of datasheets across the broader ML community to ensure the responsible and effective use of sensor data.

"Paradoxically, data is the most under-valued and de-glamorised aspect of AI" -- Google research authors of "Data Cascades in High-Stakes AI." "Data is food for AI" -- Andrew Ng, UC Berkeley professor and pioneer of the data-centric AI philosophy. Machine learning has seen a bifurcation towards both smaller and larger models in recent years. Large-scale language models with hundreds of billions of parameters are being released regularly, and, with no signs of performance saturation, we can expect to see this trend continue. On the flip side, the field of tiny machine learning (TinyML) -- deploying machine learning models on resource-constrained microcontrollers -- is also starting to take hold. Commercial applications for TinyML already exist, from keyword spotting in smartphones (e.g., "Hey Siri" and "OK Google") to person detection for controlling intelligent lighting, HVAC, and security systems.

This system, created by a team at St. Andrews University in Scotland, uses radar technology and Google's Project Soli to identify a wide variety of items. In the future, when you come across a strange thing or material, all you'll need to do is give it a scan before promptly receiving an answer.