Introduction

History is often told as a smooth arc of progress, a steady march from ignorance to enlightenment, from primitive tools to powerful machines. Artificial intelligence did not unfold that way. Learn more...

Can Machines Think?

Date: 1950

Why it mattered: The question "Can machines think?" was famously posed by Alan Turing, a British mathematician and pioneer in computer science, in his seminal 1950 paper, "Computing Machinery and Intelligence". Turing reframed the question into a practical test, now known as the Turing Test, to determine whether a machine exhibits intelligent behavior indistinguishable from that of a human.

Inflection point: The Turing Test remains a benchmark in AI research.

The far side (humor): Alan Turing publishes "Computing Machinery and Intelligence" and casually asks the most terrifying question in history: "Can machines think?" The entire academic world stares at him like he just suggested dating his toaster. Turing: "I mean - hypothetically - if a machine could fool you into thinking it's human..." Everyone else: "Alan, blink twice if the machine is holding you hostage."

The Dartmouth Conference

Date: 1956

Why it mattered: Before Dartmouth, there was no field of AI: only scattered ideas. John McCarthy coins the term "Artificial Intelligence." Researchers unite symbolic reasoning, learning, and computation. AI becomes fundable, teachable, and institutionalized.

Inflection point: Naming something makes it real and worthy of investment.

The far side: The name stuck like glue on a pizza. A bunch of nerds (McCarthy, Minsky, Shannon, etc.) meet in New Hampshire and declare: "We're gonna solve intelligence in one summer!" They name the field "Artificial Intelligence." Spoiler: It took a little longer than one summer. The conference ends with everyone optimistic and zero actual results. Like every startup pitch deck ever.

Rosenblatt's Perceptron

Date: 1958

Why it mattered: This was the first learning algorithm inspired by the brain. It introduced machine learning as learning from data. It sparked optimism that machines could learn autonomously. It triggered the first AI hype cycle... and later, backlash.

Inflection point: Intelligence could be learned, not hard-coded.

The far side: Frank Rosenblatt builds the Perceptron - a single-layer neural net that can learn patterns. The New York Times runs a headline: "The Navy reveals the embryo of an electronic computer today that it expects will be able to walk, talk, see, write, reproduce itself, and be conscious of its existence." Reality: It could barely classify dots. Hype level: 11 out of 10. Perceptron becomes the first AI celebrity... until it gets canceled hard.

ELIZA Talks Back

Date: 1966

Why it mattered: Joseph Weizenbaum's ELIZA showed that humans emotionally respond to machines. ELIZA was the first conversational AI. Users formed attachments despite knowing it was a simple pattern matching application. It raised early ethical concerns about AI deception.

Inflection point: AI wasn't just about thinking; now, it was also about relationships.

The far side: Joseph Weizenbaum creates ELIZA - a simple chatbot that pretends to be a therapist by rephrasing everything you say as questions. User: "I'm sad." ELIZA: "Why do you feel sad?" People lose their minds. One woman asks Weizenbaum if she can have private sessions with ELIZA. Weizenbaum (horrified): "It's just pattern matching!" Everyone else: "No, it's my friend."

The First AI Winter

Date: 1974-1980

Why it mattered: Reality collided with hype. Government funding collapsed after unmet promises. Governments pulled support from universities. Researchers learned humility and rigor.

Inflection point: AI learned to survive disappointment.

The far side: Perceptrons get mathematically proven limited (Minsky/Papert book). Congress: "Where's my thinking robot?" Researchers: "Uh, working on it?" Funding dries up. Everyone goes quiet for a decade. AI learns its first lesson: Don't overpromise.

Deep Blue Defeats Garry Kasparov

Date: 1997

Why it mattered: The first time a machine beat a reigning world champion at chess. Symbolic AI triumphs over human intuition. It proved that brute-force plus heuristics could outperform genius. The public perception of machine intelligence changed overnight.

Inflection point: Machines could surpass humans in elite cognition.

The far side: IBM's chess computer beats the world champion. Kasparov storms off the stage muttering about "alien pawns" and "the computer smelled funny." Deep Blue, in binary: "Bro. I don't even have a nose. And I get doused with cologne."

Hinton Revives Deep Learning

Date: 2006

Why it mattered: Neural networks were nearly abandoned, then suddenly resurrected. Deep belief networks make multi-layer learning feasible. Academic AI pivots back to neural approaches, setting the stage for modern deep learning.

Inflection point: Old ideas + new math + more compute = breakthrough.

The far side: Geoffrey Hinton (and others) figures out backpropagation (how to train multi-layer neural nets). The field gets a second wind. Hinton becomes the quiet godfather who everyone ignores for 20 years... then suddenly worships.

AlexNet Wins ImageNet

Date: 2012

Why it mattered: This is the true beginning of the modern AI boom. AlexNet produced a 10x error rate reduction over previous vision systems. GPU-accelerated deep learning proves superior, and Big Tech pivots overnight.

Inflection point: AI stops being theoretical and starts being dominant.

The far side: Geoffrey Hinton's team (Hinton, Krizhevsky, Sutskever) drops AlexNet and suddenly computers can see cats better than your mom can. The entire computer-vision community collectively screams "WE'RE ALL DOOMED... but also look at the cute-kitten detection accuracy!"

Google Acquires DeepMind

Date: 2014

Why it mattered: AI becomes a strategic corporate asset, not just for research. Talent becomes the most valuable resource. Reinforcement learning gains massive funding. AI labs become central to Big Tech strategy.

Inflection point: AI moves from academia to industrial scale.

The far side: Google buys DeepMind for half a billion dollars. DeepMind's big demo? AlphaGo isn't even out yet. Everyone: "Why did you spend that much on a game AI?" Google: "Trust us." Everyone: "We've heard that before!"

AlphaGo Defeats Lee Sedol

Date: 2016

Why it mattered: The first AI victory based on intuition, not brute force. Machines discover strategies humans never imagined. China launches a national AI strategy shortly after the match. AI becomes geopolitically significant. The race is on!

Inflection point: AI begins to exceed human creativity.

The far side: DeepMind's AlphaGo destroys the Go world champion 4-1. Lee Sedol, in post-match interview: "I felt like I was playing against God, and God was trash-talking me in binary." AlphaGo, silently: "gg ez"

Jensen Huang Donates DGX-1 to OpenAI

Date: 2016

Why it mattered: Gave OpenAI unprecedented compute access. Enabled training of large-scale language models. Cemented NVIDIA as the backbone of AI. Led to the development of ChatGPT.

Inflection point: Compute becomes the primary limiting factor and NVIDIA controls it.

The far side: NVIDIA CEO Jensen Huang personally hand-delivers the very first DGX-1 (the "AI supercomputer in a box") to OpenAI's office. Signs it: "To the future of humanity." Sam Altman and team use it to kickstart early GPT work. Jensen basically says: "Here's the shovel for your gold rush. You're welcome." He'll do it again in 2025 with DGX Spark. Old habits die hard.

Attention Is All You Need

Date: 2017

Why it mattered: One paper changes everything. Introducing transformers. Enabled scaling of language models. Made GPT, BERT, and ChatGPT possible.

Inflection point: Architecture, not data alone, unlocks intelligence.

The far side: Eight Google researchers drop the Transformer paper. Title sounds like a Zen koan. Key line: "Attention Is All You Need." The entire field reads it and collectively screams: "OH MY GOD THIS IS IT." Foundation of every modern LLM. Quietly the most influential paper of the decade, perhaps the century.

GPT-3 Shows General Language Ability

Date: 2020

Why it mattered: For the first time, a model feels general. Writes, codes, reasons, explains. Few-shot learning emerges. AI becomes a creative partner.

Inflection point: AI stops being narrow.

The far side: OpenAI drops GPT-3 (175 billion parameters). It can write essays, code, poetry, and fake news so convincingly that people start asking: "Is this the beginning of the end?" Internet: "Finally, an AI that can pass my English exam!" Also Internet: "Finally, an AI that can write my breakup texts and resignation letter!"

ChatGPT Goes Public

Date: 2023

Why it mattered: AI becomes a mass consumer product. 100M users in weeks. Triggers global regulatory response. Forces every company to rethink workflows. AI becomes a household name.

Inflection point: AI becomes infrastructure and part of daily life.

The far side: OpenAI quietly drops ChatGPT. Three days later: Every student on Earth has an A+ essay. Every Twitter thread is now 80% GPT-generated roasts. Your grandma is asking it for cookie recipes and getting philosophical treatises on butter.

AI Safety Enters the Conversation

Date: 2023-Present

Why it mattered: Power finally triggers restraint. Frontier model regulations emerge. Red teaming and alignment research take off. AGI risk enters mainstream policy.

Inflection point: Intelligence is no longer just a technical problem: it's a governance one.

The far side: AI agents are booking vacations without permission. Robot vacuums are unionizing. Your smart fridge is judging your yogurt choices. And somewhere in a garage, a guy is still using two GTX 580s to train a tiny LLaMA model and calling it "retro AGI." The end of the world is feared near.