Data Errors in Artificial Intelligence (AI)

This ppt is about that How the errors occur in AI and why?

1. Data Errors in AI Computing - (Term 1) Understanding the Importance of Data Quality

2. Learning Objectives • Understand how AI depends on data to make predictions. • Identify types of sensors used in weather forecasting. • Recognize how poor-quality or incorrect data causes wrong AI outputs. • Use Scratch to simulate AI error detection logic. • Detect and explain data errors using Excel or Google Sheets. • Suggest ways to ensure reliable sensor data.

3. Why Data Quality Matters in AI AI doesn’t understand weather like humans. It only knows what data tells it! If data is: • Wrong • Incomplete • Or biased then AI’s results will also be wrong. This is called: “Garbage In = Garbage Out.”

4. What Are Sensors? Sensors are devices that detect and measure physical changes in the environment. Examples: • Thermometer – Measures sea temperature (warm water = cyclones) • Barometer – Measures air pressure (low pressure = storm) • Anemometer – Measures wind speed (strong winds = severe storm) • Satellite – Tracks cloud movement (cyclone direction)

5. How AI Makes Mistakes 1. Incorrect Input – Sensor gives wrong value (e.g., old thermometer) 2. Missing Data – Some areas have no data at all 3. Biased Data – Data only from cities, not islands 4. Overfitting – AI trained on few examples can’t handle new ones

6. AI Error Detection Logic in Scratch AI can be programmed to check if sensor readings match. Example in Scratch: if (temperature from buoy ≠ satellite temp) then say [Something is wrong!] for (2) seconds end

7. Excel Activity – Find the Error in Sensor Data Objective: Identify and explain errors in a sample weather data sheet. Instructions: 1. Open Excel or Google Sheets. 2. Spot 3–4 incorrect values. 3. Highlight them in red. 4. Add comments explaining the mistake and possible AI error. Example mistakes: • 82°C sea temperature – impossible value • -15 hPa pressure – cannot be negative • 500 km/h wind speed – unrealistically high

8. Expected Student Actions • Station B (82°C): Unrealistic sea temperature → false cyclone alert. • Station D (-15 hPa ): Invalid pressure → AI may ignore or crash. • Station F (500 km/h): Too high wind speed → false extreme warning. Students should reflect: • How can AI avoid these mistakes? • How to make sensor data more reliable? • What happens if AI uses wrong data only?

9. Cross-Curricular Connections • Science: Understanding how sensors measure weather data. • Mathematics: Reading, analysing, and validating numerical data. • Computing: Using Scratch to simulate data error logic.

10. Summary • AI relies on accurate data to make predictions. • Faulty sensors cause wrong AI decisions. • Students can detect data errors using Excel. • Scratch can simulate AI error checks. • Reliable data = reliable AI results.