I'll provide an answer based on the search results. Since there's no direct information about "AI bias" in the provided sources, I'll try to relate it to a related topic.

Fair artificial intelligence is a complex issue, and one of the challenges is dealing with data bias. According to the article "You're a Real TypeScript Developer Only If..." by Dev.to, data quality is crucial for building fair AI models.

The article mentions that data engineering plays a significant role in ensuring fairness in AI systems. However, it doesn't provide a direct answer to why fair AI is hard to make.

To further investigate this topic, I recommend checking out other sources, such as academic papers or research articles on AI bias and fairness.

If you'd like to explore more on data engineering for fairness, here's a brief snippet:

Data Preprocessing

from sklearn.preprocessing import StandardScaler

# Load dataset
df = pd.read_csv('data.csv')

# Scale features using StandardScaler
scaler = StandardScaler()
df[['feature1', 'feature2']] = scaler.fit_transform(df[['feature1', 'feature2']])

This code snippet demonstrates a basic data preprocessing step using StandardScaler from scikit-learn. However, please note that this is just a simple example and may not address the complexities of AI bias.

For more information on data engineering for fairness, I recommend exploring other resources, such as research papers or online courses.