Edge AI means running artificial intelligence directly on a nearby device, such as a phone, smart camera, robot sensor, or small computer, instead of sending every piece of data to a faraway cloud server. This matters because devices can react quickly, use less internet bandwidth, and keep more data private. Students can think of Edge AI as putting a small trained brain inside the tool that collects the data.
It connects computer science, statistics, and data analysis because the device must turn real-world measurements into useful predictions.
Key Facts
- Edge AI runs an AI model on a local device instead of only in the cloud.
- Inference means using a trained model to make a prediction: input data + model = output prediction.
- Latency is the delay before a result appears, often measured as latency = response time - request time.
- Accuracy can be measured as accuracy = correct predictions / total predictions.
- Bandwidth saved = data not sent to the cloud, which can reduce cost and improve speed.
- A common Edge AI pathway is sensor data -> preprocessing -> model inference -> decision -> action.
Vocabulary
- Edge AI
- Edge AI is artificial intelligence that runs on a local device near where data is collected.
- Inference
- Inference is the process of using a trained machine learning model to make a prediction from new data.
- Sensor
- A sensor is a device that measures information from the physical world, such as light, sound, motion, or temperature.
- Latency
- Latency is the time delay between sending information into a system and receiving a result.
- Cloud Computing
- Cloud computing uses remote servers on the internet to store data, run programs, or process information.
Common Mistakes to Avoid
- Thinking Edge AI means no internet is ever used. Edge AI can work locally, but many systems still use the cloud for updates, backups, or training larger models.
- Confusing training with inference. Training teaches a model using many examples, while inference uses the finished model to make predictions on new data.
- Assuming Edge AI is always more accurate than cloud AI. Accuracy depends on the model, data quality, sensor quality, and how much computing power the device has.
- Ignoring privacy just because data stays on the device. Local processing can reduce risk, but apps may still store, share, or upload results unless the system is designed carefully.
Practice Questions
- 1 A smart camera processes 30 video frames per second on the device. If each frame would be 2 megabytes when sent to the cloud, how many megabytes per second of upload bandwidth are saved by processing locally?
- 2 An Edge AI model correctly identifies 92 images out of 100 test images. What is its accuracy as a decimal and as a percent?
- 3 A school robot must stop immediately when it sees an obstacle. Explain why Edge AI might be a better choice than sending camera data to the cloud before deciding.