A
(Brief)
History
of AI_
From Alan Turing's imitation game to autonomous agents — eight decades of machines learning to think, fail, and think again. A structured overview of artificial intelligence's defining moments.
The Turing Test
Alan Turing asks "Can Machines Think?" He proposes the Imitation Game — if a machine can fool a human, perhaps it thinks. The single most influential question in AI history.
Dartmouth Conference
McCarthy, Minsky, Shannon & friends coin "Artificial Intelligence" at a summer workshop. The field is officially named. They estimate they'll crack it in one summer. Reader: they did not.
The Perceptron
Frank Rosenblatt builds the first neural network hardware. The New York Times declares it will "walk, talk, see, write, reproduce itself and be conscious." Expectations were calibrated differently then.
ELIZA — First Chatbot
Weizenbaum creates ELIZA at MIT — a program mimicking a therapist. Users form emotional attachments. Weizenbaum is horrified by their credulity. The AI therapist debate begins 60 years early.
is Named.
First AI Winter
Minsky & Papert demonstrate perceptrons cannot learn XOR. Funding collapses. The UK Lighthill Report calls AI "disappointing." DARPA withdraws. The field enters hibernation.
Expert Systems Boom
AI rebounds via rule-based Expert Systems encoding human domain knowledge. XCON saves DEC $40M per year. Lisp machines proliferate. Then the rules become impossible to maintain at scale.
Second AI Winter
Expert systems fail to scale. Lisp machine market implodes. DARPA cuts funding again. Neural networks appear dead. The field shrinks to academic enclaves. Second full collapse in 20 years.
Deep Blue Defeats Kasparov
IBM's Deep Blue defeats world chess champion Kasparov 3.5–2.5. Kasparov alleges computational cheating. IBM retires Deep Blue immediately. The optics remain unresolved to this day.
We are on the edge of change comparable to the rise of human life on Earth.
Hinton's Breakthrough
Geoffrey Hinton publishes "Learning Multiple Layers of Representation." Deep neural networks actually work. The AI community largely ignores it. The slow fuse is lit.
AlexNet — ImageNet
AlexNet surpasses ImageNet benchmarks by over 10 points. GPU acceleration meets deep learning. Every major lab pivots overnight. The modern era begins here.
GANs — Generative AI Born
Goodfellow invents GANs at 2am after a debate. Two competing networks fight each other. Generative AI is born. Creative industries begin a long, complicated reckoning.
AlphaGo — Go Defeated
DeepMind's AlphaGo defeats Lee Sedol 4-1 at Go — more board states than atoms in the observable universe. Sedol retires in 2019, citing the machine as undefeatable.
All You Need.
The Transformer Paper
Google Brain publishes "Attention Is All You Need" — eight authors, one paper. It replaces recurrent networks entirely. Every modern foundation model runs on this architecture.
BERT & GPT-1
Google releases BERT; OpenAI ships GPT-1. Language models demonstrate genuine text comprehension. The scaling arms race begins — every point of compute translates to capability.
GPT-3 — 175B Parameters
OpenAI scales to 175 billion parameters. GPT-3 writes essays, passes professional exams, generates working code. API waitlist exceeds one million developers. Microsoft invests $1 billion.
Image Generation Emerges
DALL·E, Midjourney, and Stable Diffusion make text-to-image accessible at consumer scale. Generative media proliferates. Professional illustrators face an unprecedented market disruption.
ChatGPT — Mass Adoption
OpenAI releases ChatGPT. One million users in five days. One hundred million in two months — the fastest consumer product adoption in recorded history. Google internally declares Code Red.
Proliferation
Google launches Bard → Gemini. Meta open-sources LLaMA. Anthropic releases Claude. Mistral emerges from Paris. The landscape fragments rapidly across dozens of competing foundation models.
Agents & Reasoning
The paradigm shifts from assistants to autonomous agents. Cursor writes code. Devin handles software engineering. OpenAI o1 implements chain-of-thought reasoning. DeepSeek demonstrates efficiency breakthroughs.
Open Horizon
Multimodal. Autonomous agents. Drug discovery. Scientific reasoning. Physical robotics. The AGI debate intensifies. Whether or not general intelligence arrives this decade, the economic transformation is already underway.
Agents Enter the Workforce
AI agents handle entire workflows — legal research, financial analysis, medical imaging, code review. White-collar disruption accelerates faster than policy can respond. The assistant era ends. The colleague era begins.
Multimodal Becomes Standard
Text, image, audio, video, and code collapse into unified models. AI that sees, hears, reads, and acts simultaneously becomes the baseline. The word "chatbot" is retired. Good riddance.
Scientific Discovery Accelerates
AlphaFold showed the way. The next wave extends into drug design, materials science, climate modelling, and fundamental physics. Discoveries arrive that we weren't specifically looking for.
Regulatory Frameworks Emerge
The EU AI Act is a beginning, not an end. Expect global divergence: EU tightens, US fragments by sector, China integrates AI into state infrastructure. An international body is proposed — and disputed.
Physical AI — Robots in Context
Foundation models extend into physical embodiment. Robots navigate, manipulate, and reason in unstructured environments. Warehouses, hospitals, construction. Physical labour re-prices again.
Personalised AI — One Per Human
Every person carries a persistent AI companion trained on their life: communication style, medical history, professional knowledge. Memory and continuity become design challenges more than technical ones.
AGI — The Contested Threshold
Whether a system reaches AGI depends entirely on definition. Narrow tasks: already surpassed. Common sense: rapidly closing. Flexible real-world agency at human level: the decade's open question.
The Unknown Unknowns
Every previous forecast missed the thing that actually happened. No one predicted GPT-3's emergent capabilities. No one predicted ChatGPT's adoption curve. The most significant development of the decade is probably not on any roadmap.