What are we covering ?

In this blog series (You will see more blogs on this topic)), we’re diving into the world of AI and how it solves problems—basically, how to make computers think like super-smart detectives. We’ll start with the basics, like how AI began, some fun challenges like the Turing Test (can a machine fool you into thinking it’s human?) and the Winograd Schema (spoiler: it’s about tricky sentences). Then, we’ll explore how AI searches for answers—whether it’s taking the scenic route (depth-first search) or the fastest route (breadth-first search).

We’ll also uncover how AI gets creative with shortcuts, avoids dead ends, and even uses strategies inspired by ants and evolution (yes, ants!). Whether it’s finding the best path, playing games like a pro, or planning your tasks better than you ever could, AI’s got it covered. Stick around, and you’ll see how machines solve problems smarter, faster, and sometimes in ways that’ll make you say, “Wait, what?”

History and Philosophy :-

Intelligence :-

Yep we will get into short history of AI , but what is intelligence we are talking here huuh 🤔 ?

Remembering 🧠: Learn from the past (case-based reasoning).
Understanding 🌍: Know what’s going on (logic & knowledge).
Imagining 🚀: Dream up solutions (trial, error, planning).

Basically, brains + ideas + action!"

A decade in AI :-

AI's glow-up in the last decade was wild, but let's not forget, intelligence has always been a hot topic. A lot of big thinkers had their say:

Herbert Dreyfus 🤔: Argued that intelligence is more about unconscious instincts than formal rules. So no, you can't just code your way to being a genius.
John Searle: The Chinese Room experiment asks, “If you follow rules to process Chinese, does that mean you understand it?” Spoiler: nope.
Roger Penrose: Believed there's some quantum magic in our brains that AI just can't explain...yet. 🧠✨
Alan Turing: The Turing Test – can a machine think? (Answer: "Who even knows?" 😅) Also, in 1950, Turing said the question was "too meaningless" but suggested a test to see if machines could imitate human conversation.

And here comes the fun stuff:

ELIZA (1966) 🗣️: A simple program that acted like a psychotherapist. It followed basic rules and sort of understood you. It was basically the AI equivalent of asking, "How does that make you feel?"
The Turing Test: Test machines by making them chat like humans. In 2013, the Loebner Prize (AI’s version of a talent show) had Izar answer questions like: "What's your favorite fruit?" 🍌. But, AI wasn’t great at emotions (or bananas).
Winograd Schemas 💡: An AI challenge that's harder than just Googling answers. It makes machines figure out tricky pronouns like “Who does ‘they’ refer to?” Example: “The city councilmen refused the demonstrators a permit because they feared violence.” (Is ‘they’ the councilmen or demonstrators?) Trickier than it sounds! This problem of resolving what a pronoun, or a noun phrase is known as anaphora resolution.

Now, let’s look at the neuron evolution:

Perceptron (1943) 🧠: A baby neural network for binary classification – super basic.
Minsky & Papert (1958): Called out Perceptron’s limits – couldn’t handle messy data.
1986: Backpropagation! Networks learned from their mistakes – multi-layer networks could classify non-linear data like pros.
2012 Hinton: Deep networks took over and crushed tasks like computer vision – seeing thousands of objects.

AI went from "just ok" to "next-gen cool" thanks to better algorithms, bigger data, and powerful computing power. But does AI really think? Let’s leave that question to philosophers, shall we? 😏

Minds and Machines made of matter :-

AI : The myth , the mystery

Thomas Hobbes: The OG AI Grandpa 🤖👴

Way back in the 1600s, Thomas Hobbes was like, “Thinking is just manipulating symbols,” inspired by Galileo’s idea that everything is made of particles (including our thoughts). 🧠💭

Hobbes believed animals are machines, but humans have a mind (that’s a flex). 🐾➡️🧠

Descartes 🧳: Thoughts are like symbols in algebra. Not the world, just how we think about it! He said, Reasoning = Computation: Adding, subtracting, or just thinking is like math. 😎

The Paradox of Mechanical Reason: If thinking is symbol manipulation, who’s doing the thinking? Can a machine understand meaning? 🤯 It gets even weirder when you think of a "little man" (the homunculus) inside the machine, guiding it—kind of like a mini brain inside the machine. 🧑‍🔬

And then, we get into mythical creations! 🏛️

Hephaestus made Talos (robot guard). 🦾
Pygmalion created Galatea (ivory woman-turned-real!). 🏺✨
Paracelsus made a humunculus (little man). 🧑‍🔬

AI has always had big dreams... just with more mechanical twists than we'd like to admit. 😅

The Real Mechanisms: From Ducks to Computers 🦆💻

Let’s fast forward from myth to real-world machines that got the ball rolling in AI!

Vaucanson’s Duck (1738) 🦆: A mechanical marvel that could eat, digest, and even poop! (A duck with a job!)
Von Kempelen’s Chess Playing Turk (1770s) ♟️: A “chess-playing robot” that wasn’t really a robot—spoiler, there was a human inside. Classic prank!
Mechanical Arithmetic 🔢: Pascal’s Calculator (1642) – The first step towards calculators. It could add and subtract—just don't ask it to do your taxes.
Stepped Reckoner (1672) 🔄: Leibniz's upgrade on Pascal’s machine, could do addition, subtraction, multiplication, and division. Fancy, right?
Leibniz’s Calculus Ratiocinator 📐: A system to help solve math problems, but it’s basically a precursor to computer logic!
Commercial Success (Arithmometer) 💸: Thomas de Colmar made the first commercially successful adding machine in 1820. It added up to your wallet too!
The First Computer (1837) 🖥️: Charles Babbage designed the first mechanical computer, but didn’t quite finish it. Story of every project, huh?
Babbage's Difference Engine ➗: A machine to calculate and print tables—way too advanced for its time (but it worked!).
Jacquard Looms (1801) 🧵: Not AI, but Joseph Marie Jacquard’s loom used punch cards to automate weaving—basically the grandparent of computer programming.
The Analytic Engine (1837) 🔧: Babbage’s dream machine, capable of any calculation, if only he’d finished it. Enter Ada Lovelace, the first computer programmer 💻!
ENIAC (1945) ⚡: The first true electronic computer. Big, fast, and noisy. Think of it as the granddaddy of modern computers—minus the Wi-Fi.

From mechanical ducks to calculating machines, humans were definitely thinking ahead! 🤯

Modern History of AI :-

The Dartmouth Conference & Early AI Pioneers 💡

In 1956, John McCarthy, Marvin Minsky, and Claude Shannon hosted the legendary Dartmouth Conference—the birthplace of AI! 🤖 Their big idea? Every aspect of intelligence could be described so precisely that machines could mimic it.

The Organizers:

John McCarthy: The man who named AI and created the Lisp programming language for AI geeks to love. 🖥️
Marvin Minsky: Co-founder of the MIT AI Lab, frames theorist, and AI legend. Also wrote The Emotion Machine (yes, AI has feelings!). 💭
Claude Shannon: The “Father of Information Theory” who helped bring McCarthy and Minsky together at Bell Labs. 📡
Nathaniel Rochester: The IBM engineer behind the first assembly language and a checkers-playing program (too bad IBM got freaked out by “electronic brains” 😱).

The Show Stealers:

Herbert Simon & Alan Newell: They weren’t the most famous folks at first, but they stole the show with the Logic Theorist – the first program to prove theorems like a human! 🤓 It even found elegant proofs in Principia Mathematica. 🏆

The Info-Processing View:

Simon & Newell’s General Problem Solver (GPS) was a game-changer! It used heuristics and means-ends analysis, kind of like how we solve problems (but, you know, with less coffee ☕).
CMU’s SOAR by John Laird expanded this approach, turning the info-processing method into a solid AI framework.

AI was born with big brains and even bigger ideas—and it all started at Dartmouth! 🎓

AI : Problem solving

Problem Solving in AI: Search Methods 🧠🔍

AI loves a good challenge, and problem-solving is at the core of it all! Let's explore how search methods and reasoning are used to solve puzzles, real-world tasks, and more.

Autonomous Agent on a Football Field ⚽🤖

Imagine an autonomous agent (like a robot player) on a football field. It has to move, kick the ball, and dodge players—all while making decisions in real-time! To make this happen, the agent needs to make assumptions:

The world is static
The world is completely known
Only one agent changes the world
Actions never fail

Representation of the world is taken care of (to start with at least)

Search vs. Knowledge: Solving the Rubik’s Cube 🧩

Imagine solving the Rubik's Cube. It’s a combination of search and knowledge:

Search: Trying out different moves to reach the solved state.
Knowledge: Knowing the algorithms or patterns to make moves efficiently.

A brute-force approach? Try every possible combination (good luck 😅). But, using algorithms? A lot faster!

Search vs. Reasoning: Solving Sudoku 🧩

Solving Sudoku is like solving a mystery:

Search: Explore possible numbers to fill the grid.
Reasoning: Use logic to eliminate impossible number choices based on constraints.

You can use search when stuck, but reasoning will help you solve it like a pro. 🧠

A Map Colouring Problem 🌍🎨

The map colouring problem: How to color a map with as few colors as possible, so neighboring regions don’t share the same color?

Search: Try coloring the regions in every possible way.
Reasoning: Use logical rules to minimize the number of colors and avoid conflicts.

AI takes both search and reasoning to figure out the best solution—whether it’s solving puzzles, playing football, or coloring maps! 🎮

An Autonomous Agent :-

A diagram of an autonomous agent

Alright, picture an autonomous agent as a super-cool robot that’s basically the AI version of that friend who never needs help, always figures things out, and has a backup plan just in case things go sideways.

Signal Processing: It’s like the robot’s ears and eyes rolled into one—it senses the world (with sensors, like a superhero scanning for danger).
Neuro-Fuzzy Reasoning: When things get blurry or unclear (like when you’re deciding if you should nap or work), the robot uses “neuro-fuzzy” thinking to make decisions, even if it doesn’t have all the facts.
Symbolic Reasoning: This is the robot using symbols and logic like an old-school chess master to figure out its next move, just like you’d solve a Rubik's Cube (well, almost as quickly).
Planning: The robot’s like, “I’ll plan out all my moves ahead of time!”—like a super organized person who maps out their day to avoid awkward situations.
Search & Problem Solving: It’s basically playing hide-and-seek with its surroundings, searching for the best way to get to its goal without running into walls, people, or other obstacles.
Adversarial Reasoning: And when something tries to stop it, like another robot or a sneaky obstacle, the agent thinks “Game on!” and uses clever tactics to outsmart its foe.

So, this autonomous agent is like the ultimate problem solver—smart, sneaky, and always ready to take on the world without breaking a sweat!

So what exactly are we discussing from here-on ?

In the image you can see that their are two verticals for architecture of problem solving . We will skip the “Knowledge and Experience” vertical for now and focus on “First principles” which includes , representation in logic , constraint processing, logical deduction and searching.

AI : Search Methods and Problem Solving

Table of contents