AI App Development For Beginners

Overview

When you ask a chatbot a question and it answers back, what’s actually happening under the hood? This section defines artificial intelligence (AI) in plain language, gives a quick history, and maps the main types of AI-so you can recognize them in products and APIs. It also ties the concepts to the AI Map Framework so you can categorize a system by behavior.

Quick Reference

AI Map Framework (how to classify an AI system)

• Input: What data it consumes (text, images, events)
• Goal: What output it produces (label, prediction, response)
• Method: Rules, ML model, or generative model
• Limits: What it can’t reliably do (hallucinations, missing context, bias)

Parameters

This chapter also includes an API-ready way to think about “AI types” as parameters you’ll see in developer guides.

Code Example

Below is a minimal Python example that uses the same classification idea from the AI Map Framework: choose `ai_mode` and `task_type`, then call an API that returns structured results.

import os
import requests

API_URL = "https://api.example.com/v1/ai"  # Replace with your provider endpoint

def run_ai(task_type: str, ai_mode: str, prompt_or_input: str, max_tokens: int = 200):
    payload = {
        "task_type": task_type,          # "classification" | "prediction" | "generation"
        "ai_mode": ai_mode,              # "rule_based" | "machine_learning" | "generative"
        "prompt_or_input": prompt_or_input,
        "max_tokens": max_tokens,
        "return_format": "json",
    }

    headers = {
        "Authorization": f"Bearer {os.environ['AI_API_KEY']}",
        "Content-Type": "application/json",
    }

    resp = requests.post(API_URL, json=payload, headers=headers, timeout=30)
    resp.raise_for_status()
    return resp.json()

if __name__ == "__main__":
    # Nina-focused concept mapping: generative mode for response generation
    result = run_ai(
        task_type="generation",
        ai_mode="generative",
        prompt_or_input="Explain what AI is in one paragraph.",
        max_tokens=120,
    )
    print(result)

Response Format

Expected JSON structure (typical for API responses that support structured outputs):

{
  "id": "req_12345",
  "task_type": "generation",
  "ai_mode": "generative",
  "output": {
    "text": "string result",
    "confidence": 0.0
  },
  "usage": {
    "input_tokens": 0,
    "output_tokens": 0
  },
  "error": null
}

Field notes:

• `task_type`: mirrors the parameter you sent (`classification`, `prediction`, `generation`)
• `ai_mode`: mirrors the parameter you sent (`rule_based`, `machine_learning`, `generative`)
• `output.text`: generated or summarized text
• `output.confidence`: may be omitted depending on the endpoint
• `usage`: token counts for cost and limits tracking
• `error`: null on success; an object on failure

Notes & Best Practices

• Rate limits: Use fewer requests by batching inputs when possible; check provider docs for `429` handling.
• Error handling: Always call `resp.raise_for_status()` (or check status codes) and log `resp.text` for debugging.
• Edge cases: Generative AI may return plausible but incorrect statements; treat outputs as drafts unless your task has verifiable sources.
• Beginner mistake to avoid: Mixing “rule-based” expectations with “generative” outputs-if you need exact labels, prefer `classification` with `machine_learning` instead of free-form generation.

...