90 Years of Computing History

The Epic of Computing
From Turing's Infinite Tape to Google Willow and OpenAI System-2

In 1936, Alan Turing conceived the universal engine of computation. Nine decades later, humanity achieved below-threshold quantum error correction with Google Willow and deliberate System-2 reasoning with OpenAI. This is the complete trajectory of mechanical intelligence.

1936
ERA 1 • 1936 – 1950 \"Can machines think?\" — Alan Turing, 1950

The Mathematical Oracle & The Infinite Tape

Artwork generated with Gemini
Chapter narration
0:00

In 1936, in the revolutionary paper \"On Computable Numbers\", British mathematician Alan Mathison Turing formulated the Universal Turing Machine. He proved that a single hypothetical device, reading and writing symbols on an infinitely long paper tape, could simulate any conceivable calculating machine.

During WWII at Bletchley Park, Turing's theoretical work became operational. He designed the Bombe to break German Enigma cipher transmissions, while Tommy Flowers built Colossus (the first electronic digital computer with 1,500 vacuum tubes). In 1950, Turing published the seminal \"Computing Machinery and Intelligence\", introducing the Turing Test.

Interactive Simulator: The Turing Machine Tape
Emulator Active
▼ Estado: q0
Rule: Invert bits 0 ↔ 1 and advance head right.
01 Formal split between Hardware (machine) and Software (program on tape).
02 Bletchley Park saved millions of lives by automating military cryptanalysis.
03 Philosophical origin of the Turing Test and the field of Artificial Intelligence.
1947
ERA 2 • 1945 – 1960 \"Any problem that can be computed can be held in memory.\" — John von Neumann

Von Neumann Architecture & The Solid-State Transistor

Artwork generated with Gemini
Chapter narration
0:00

The first giant electronic computers, like the ENIAC (1946) with 18,000 vacuum tubes, filled entire rooms and failed constantly from heat. In 1945, John von Neumann published his EDVAC report, defining the Stored-Program Architecture: data and instructions share the same main memory, processed by a Control Unit and ALU. This structure governs almost all computers to this day.

On December 16, 1947, at Bell Laboratories, John Bardeen, Walter Brattain, and William Shockley invented the point-contact germanium transistor — a solid-state switch with no moving parts, cool, tiny, and infinitely faster than vacuum tubes. In 1956 at Dartmouth, John McCarthy officially coined the term \"Artificial Intelligence\".

Physical Comparison: Vacuum Tube vs. Silicon Transistor
Vacuum Tube (1945)
Size: ~10 to 15 cm
Power: ~5 to 10 watts
Reliability: Burns out in ~2,000 hours
Speed: ~10 to 100 kHz
VS
Solid-State Transistor (1947+)
Size: From mm to 2 nanometres
Power: Picowatts (a billion times less)
Reliability: Decades of service life
Speed: Gigahertz (GHz)
1971
ERA 3 • 1960 – 1980 \"The number of transistors will double every two years.\" — Gordon Moore, 1965

Integrated Circuits, Moore's Law & The Intel 4004

Artwork generated with Gemini
Chapter narration
0:00

With the invention of the integrated circuit by Robert Noyce and Jack Kilby, multiple transistors could be printed on a single silicon wafer. In 1965, Gordon Moore formulated the most profitable empirical prediction in history: Moore's Law.

In 1971, Intel launched the Intel 4004: the first commercial single-chip microprocessor with 2,300 transistors. Concurrently, Bell Labs birthed UNIX and the C language (Dennis Ritchie and Ken Thompson), while ARPANET made its first transmission in 1969, weaving the foundations of the global digital infrastructure.

Moore's Law Visualizer: Exponential Transistor Growth
Ano: 1971
Chip: Intel 4004
Transistores: 2.300
1984
ERA 4 • 1980 – 2000 \"This is for everyone.\" — Sir Tim Berners-Lee

Personal Computers, The World Wide Web & Backprop

Artwork generated with Gemini
Chapter narration
0:00

Computers finally left corporate data centers and invaded homes. The IBM PC (1981) and iconic Apple Macintosh (1984) popularized the graphical interface with windows, icons, and the mouse. In 1989, at CERN, Tim Berners-Lee invented the World Wide Web (HTTP, HTML, and URLs), connecting humanity in a single hypertext web.

In AI, after a long \"winter\", David Rumelhart, Geoffrey Hinton, and Ronald Williams published the Backpropagation algorithm in 1986, proving that multilayer neural networks could learn complex internal representations through gradient descent.

Backpropagation Simulator: Synaptic Error Weight Adjustment
Input
X1: 0.8
X2: 0.2
w1: 0.500 w2: 0.350
Output
Y: 0.473
Loss: 0.2769
Loss →
2012
ERA 5 • 2000 – 2016 \"Deep learning works because the world is compositional.\" — Yann LeCun

Cloud Computing, CUDA GPUs & The Deep Learning Big Bang

Artwork generated with Gemini
Chapter narration
0:00

At the turn of the millennium, computation migrated to elastic data centers with cloud computing (AWS EC2). Concurrently, NVIDIA launched the CUDA platform in 2006, unlocking massively parallel GPU power for linear algebra and matrix operations.

In October 2012, Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton won ImageNet with AlexNet, crushing classical computer vision methods and launching the modern Deep Learning boom. In March 2016, AlphaGo shocked the world by defeating Lee Sedol at Go.

2017
ERA 6 • 2017 – 2024 \"Attention is all you need.\" — Vaswani et al., Google Brain

The Transformer Revolution & The Generative AI Boom

Artwork generated with Gemini
Chapter narration
0:00

In June 2017, eight Google researchers published the historic paper \"Attention Is All You Need\". The Transformer architecture replaced recurrent networks with the Self-Attention mechanism, enabling supercomputers to train in parallel on billions of web pages without sequential bottlenecks.

The leap with GPT-3 (2020) and the global launch of ChatGPT (November 2022) turned AI into the fastest-adopted technology product in human history. Focus rapidly expanded to native multimodality (vision, audio, and code unified) with GPT-4o and Gemini 1.5.

2026
ERA 7 • 2024 – 2026 (THE CURRENT FRONTIER) \"We crossed the quantum threshold and the deliberate-reasoning barrier.\" — Make AI Better

The Dual Frontier: Quantum Leap (Google Willow) & System-2 Reasoning (OpenAI)

Artwork generated with Gemini
Chapter narration
0:00

In 2026, computing history reaches its most spectacular simultaneous convergence between hardware physics and cognitive intelligence:

Google Quantum AI

The Willow Quantum Chip (Dec 2024–2026)

With 105 superconducting transmon qubits, Willow became the first chip in history to cross the quantum error-correction threshold (\"below-threshold\"). As the surface-code lattice grows, the logical error rate falls exponentially.

Compute time (RCS): < 5 minutes
Classical supercomputer: 10 septillion ($10^{25}$) years
Verifiable advantage: 13,000x faster (Quantum Echoes)
Willow Superconducting Qubit Lattice Simulator:
Logical error rate: 0.000012% (below threshold)
OpenAI Frontier Reasoning

The System-2 Reasoning Revolution (o1, o3, GPT-6 Astra)

The end of the impulsive next-token guess. The 2026 frontier is governed by test-time compute scaling: dynamically allocating thinking time so the model explores hypothesis trees and corrects itself before answering.

AIME 2026 olympiad: 96.8% accuracy
SWE-bench Verified: 81.9% on autonomous engineering
Core mechanism: Process Reward Models (PRMs)
Real-Time System-2 Deliberation Tree:
Mathematical proof (AIME 2026)
✓ Step 1: Problem formalised in polynomial space
✗ Hypothesis 2A: Weak induction failed — rejected by PRM
✓ Hypothesis 2B: Bijection proved in Lean 4 (verified)
✓ Step 4: Answer verified. AIME score: 100%

The Convergence of Two Epochal Trajectories

In 1936, Turing formalized computation. In 2026, Google Willow conquers quantum error correction below threshold while OpenAI unlocks step-by-step verified reasoning. The mission of Make AI Better is to ensure this immeasurable power remains transparent, safe, and accessible to all humanity.

Scroll sideways or use ← → to travel 90 years