What is AI?10 min read

What is AI?

A robot waiter, a chess engine, and Siri walk into a bar — what makes them intelligent?
scope:Foundationaldifficulty:Beginner

The Robot Waiter

Picture this: you walk into a restaurant and a robot rolls up to take your order. It recognizes your face, remembers you like extra cheese, navigates around chairs without crashing, and even cracks a joke about the soup of the day.

Is it intelligent? It sure looks intelligent. But is it thinking? Does it understand what cheese is, or is it just following very clever instructions?

This question — what counts as intelligence in a machine — is the beating heart of Artificial Intelligence.

A Brief History: From Turing to Today

The story of AI starts with a simple question posed by Alan Turing in 1950: "Can machines think?" He proposed the Turing Test — if a human can't tell whether they're chatting with a person or a machine, the machine might as well be called intelligent.

In 1956, a group of researchers at Dartmouth College coined the term "Artificial Intelligence" and boldly predicted that machines would match human intelligence within a generation. Spoiler: they were a bit optimistic.

Here's the timeline in a nutshell:

  • 1950s-60s — The dreamers. Early programs that could play checkers and prove math theorems. Excitement was through the roof.
  • 1970s-80s — The AI Winter. Funding dried up when early promises didn't pan out. Turns out, intelligence is harder than anyone thought.
  • 1990s-2000s — The quiet comeback. IBM's Deep Blue beat chess champion Garry Kasparov in 1997. Machine learning started getting serious.
  • 2010s-now — The explosion. Deep learning, massive datasets, powerful GPUs. AlphaGo beats a Go grandmaster. ChatGPT writes essays. Self-driving cars roam the streets.

We're living in the most exciting era of AI in history. And it all started with one question: can machines think?

The Three Pillars: Perceive, Reason, Act

Every intelligent system — human or machine — does three things:

  • Perceive — Take in information from the world. For humans, that's eyes, ears, touch. For AI, it's cameras, microphones, sensors, and data feeds.
  • Reason — Make sense of that information. Find patterns. Draw conclusions. Decide what matters.
  • Act — Do something useful with that reasoning. Move, speak, recommend, predict, or create.

Our robot waiter perceives through its cameras and sensors ("there's a table with two people"). It reasons using its programming ("they haven't ordered yet, and one of them is a regular who likes pizza"). And it acts by rolling over and suggesting the margherita.

Different AI systems focus on different pillars. A computer vision system is mostly about perception. A recommendation engine (like Netflix suggesting shows) is mostly about reasoning. A robot arm in a factory is mostly about action.

The holy grail? A system that does all three as well as a human. We're not there yet — but we're getting closer every year.

A Simple AI Agent

class SimpleAIAgent:
    """A basic AI agent that perceives, reasons, and acts."""
    
    def __init__(self, name):
        self.name = name
        self.memory = {}  # remembers past interactions
    
    def perceive(self, input_data):
        """Take in information from the environment."""
        print(f"[{self.name}] Perceiving: {input_data}")
        return input_data
    
    def reason(self, observation):
        """Process information and make a decision."""
        # Simple rule-based reasoning
        if "hungry" in observation.lower():
            decision = "recommend food"
        elif "lost" in observation.lower():
            decision = "give directions"
        else:
            decision = "ask for more info"
        
        print(f"[{self.name}] Reasoning: {decision}")
        return decision
    
    def act(self, decision):
        """Take action based on the decision."""
        actions = {
            "recommend food": "How about the margherita pizza?",
            "give directions": "The exit is to your left!",
            "ask for more info": "Could you tell me more?"
        }
        response = actions.get(decision, "I'm not sure what to do.")
        print(f"[{self.name}] Acting: {response}")
        return response
    
    def run(self, user_input):
        """The full perceive-reason-act loop."""
        observation = self.perceive(user_input)
        decision = self.reason(observation)
        return self.act(decision)


# Create and run our agent
waiter = SimpleAIAgent("RoboWaiter")
waiter.run("I'm hungry, what's good?")
print()
waiter.run("I'm lost, where's the exit?")
Output
[RoboWaiter] Perceiving: I'm hungry, what's good?
[RoboWaiter] Reasoning: recommend food
[RoboWaiter] Acting: How about the margherita pizza?

[RoboWaiter] Perceiving: I'm lost, where's the exit?
[RoboWaiter] Reasoning: give directions
[RoboWaiter] Acting: The exit is to your left!
Note: AI Winters and Summers: AI research has gone through dramatic cycles of hype and disappointment. The 1970s and late 1980s saw "AI winters" where funding evaporated because early systems couldn't live up to grand promises. But each winter was followed by a warmer summer. Today's AI boom is powered by three things that previous generations didn't have: massive data, cheap computing power, and better algorithms. Will this summer last? History says to stay cautiously optimistic.

So... Is Our Robot Waiter Actually Intelligent?

Depends on who you ask! There are two schools of thought:

  • Weak AI (Pragmatists) — It doesn't matter if the robot "truly" understands anything. If it behaves intelligently and produces useful results, that's AI enough. This is the view most engineers hold.
  • Strong AI (Philosophers) — Real intelligence requires genuine understanding and consciousness. A machine that just pattern-matches isn't truly intelligent, no matter how convincing it looks.

For practical purposes, we define AI as: any system that can perform tasks that would normally require human intelligence. That includes recognizing faces, understanding speech, playing games, translating languages, driving cars, and writing code.

Notice we didn't say the system has to think. It just has to perform. That's a crucial distinction — and it's what makes modern AI so powerful and so controversial at the same time.

Quick check

What was the Turing Test designed to evaluate?
Challenge

Continue reading