Introduction to AI & Generative AI

Understand the distinction between traditional AI, Generative AI, and how a data scientist leverages both for analysis and content creation.

What it is

Artificial Intelligence (AI) is a broad field focused on creating systems that perform tasks requiring human-like intelligence. Within this, Machine Learning (ML) allows computers to learn from data without explicit programming. Generative AI (GenAI) is a subset of ML specifically designed to create new content—such as text, images, or code—by learning patterns from vast datasets. For a data scientist, traditional AI/ML models are typically used for prediction and classification, while GenAI is used for augmentation, summarization, and synthetic data generation.

Why it matters

• Accelerated Coding: GenAI tools can generate boilerplate Python code, SQL queries, or regex patterns, reducing development time.
• Data Augmentation: GenAI can create realistic synthetic data to train models when real-world data is scarce or sensitive.
• Natural Language Interfaces: Allows non-technical stakeholders to query databases using plain English via Large Language Models (LLMs).
• Automated Reporting: Summarizes complex statistical findings into readable executive summaries.

Syntax or steps

To interact with a GenAI model programmatically, you typically use an API client. The general flow involves: 1) Initializing the client with credentials, 2) Constructing a prompt (the instruction), and 3) Sending the request to receive a completion. While specific libraries vary, the conceptual pattern remains consistent across providers like OpenAI, Anthropic, or local LLMs.

Example

This example uses a generic pseudocode structure common in many Python SDKs to demonstrate how a data scientist might ask an LLM to explain a dataset's schema.

import os
# Assume 'openai' library is installed and API key is set in environment
from openai import OpenAI

client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY"))

def explain_schema(table_name, column_info):
    prompt = f"""
    You are a senior data scientist. 
    Explain the purpose of the table '{table_name}' based on these columns: {column_info}.
    Keep it under 50 words.
    """
    
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": prompt}]
    )
    
    return response.choices[0].message.content

# Usage
schema_desc = explain_schema("sales_transactions", ["id", "date", "amount", "customer_id"])
print(schema_desc)

Part-by-part explanation:

• client = OpenAI(...): Initializes the connection to the AI service using secure environment variables.
• prompt = ...: Defines the context and task. Specificity here ("under 50 words") controls output quality.
• client.chat.completions.create(...): Sends the request. The model parameter selects the specific AI engine.
• response.choices[0]...: Extracts the generated text from the structured JSON response returned by the API.

Common mistakes

• Vague Prompts: Asking "Analyze this" yields poor results. Always specify role, task, constraints, and format.
• Ignoring Hallucinations: GenAI can confidently state false facts. Always verify critical outputs against ground-truth data.
• Hardcoding API Keys: Never commit keys to version control. Use environment variables or secret managers.
• Over-reliance on Default Parameters: Adjusting temperature (creativity vs. determinism) is crucial for different tasks.

When to use it

Practice

Guided Exercise: Modify the example above to change the prompt so it generates a SQL query instead of an explanation. Hint: Change the role to "SQL Expert" and provide sample data values.

Challenge: Write a function that takes a list of customer feedback strings and uses GenAI to categorize them into "Positive," "Negative," or "Neutral." Expected Output: A dictionary mapping each string to its category.

Quick check

Q: Why is temperature setting important in GenAI applications?

A: It controls randomness. Low temperature (0.0-0.3) makes outputs deterministic and factual (good for coding/data extraction), while high temperature (0.7-1.0) encourages creativity (good for brainstorming).

Summary

Generative AI complements traditional data science by handling unstructured data and creative tasks, whereas traditional ML excels at structured prediction. Effective use requires precise prompting, rigorous validation of outputs, and understanding the probabilistic nature of LLMs versus the deterministic nature of classical algorithms.