While many data sources are predominantly software based, most ultimately rely on input originating in the physical world. Devices like smartphones and laptops are rich with sensors that capture the world around them, but increasingly everyday devices like TVs, cars, watches, doorbells, and thermostats are collecting data about the environment around us as well.
Most connected devices today incorporate some type of sensor. A device like a smartphone may contain a GPS sensor (location), an accelerometer (motion), a camera (video and still), an ambient light sensor, a touchscreen, a fingerprint sensor, a microphone, and maybe a proximity sensor. With a bluetooth or wifi connection, the list may extend to include the web of home automation and remote sensors within range. The same goes for laptops and other personal computing and communications devices. Beyond computers and phones, your car may have GPS, accelerometers, and onboard sensors monitoring braking, steering wheel position, engine performance, rain on the windshield, ambient light, tire pressure, and proximity to objects and people.
In addition to devices that you interact with directly, there are passive and active sensors in the environment around you. Satellites for weather or imagery, security cameras, traffic cameras, subway turnstiles, credit card readers, employee badge readers, connected utility meters, and a broad range of industrial and environmental monitors focused on processes as opposed to people.
While the potential for data collection from these many sources can be concerning, the reality is that most of these sensors are dedicated to a small number of applications and services that are not likely to share or combine what they access. Thus, while the subway might know that you are riding, it is unlikely that your laptop knows or cares. Data ecosystems tend to maintain closed borders, focusing on sources ands sensors relevant to their purpose and within the boundaries their business relationships define.