Sources and further reading
Every figure, quotation and case in this article comes from a public source. They are listed here by chapter, with what each one supports. Where the original is behind a paywall or no longer online, the link goes to a reliable report of it.
01 · What generative AI actually is
A logical calculus of the ideas immanent in nervous activity
The first mathematical model of a neuron.
Computing Machinery and Intelligence
The paper that proposed what is now called the Turing test.
The perceptron: a probabilistic model for information storage and organization in the brain
The perceptron, the first neural network built as a physical machine.
Perceptrons: An Introduction to Computational Geometry
The book that demonstrated the limits of single-layer networks and preceded the first AI winter.
Learning representations by back-propagating errors
The training breakthrough on the timeline.
Backpropagation applied to handwritten zip code recognition
The handwritten-digit network that became LeNet.
Gradient-based learning applied to document recognition
Describes the commercial cheque-reading systems built on LeNet.
A fast learning algorithm for deep belief nets
The result usually taken as the start of deep learning.
The transformer architecture behind today's language models.
The public release that made generative AI mainstream.
02 · The strategic landscape
Nvidia CEO Jensen Huang predicts data center spend will double to $2 trillion
The $2 trillion data-centre projection.
Gen AI: too much spend, too little benefit?
The sceptical headline card.
MIT report: 95% of generative AI pilots at companies are failing
The 95% of enterprise deployments with no measurable return, used again in chapter 03.
The revenue the AI sector would need to justify its investment.
Sam Altman, the man behind ChatGPT, is increasingly alarmed about what he unleashed
"The bad case … is, like, lights-out for all of us."
OpenAI's Sam Altman seeking trillions to fund chips for AI, report says
The reported plan to raise up to $7 trillion for chip production.
The case for AI as a transformative, and risky, technology, from the head of Anthropic.
Pause Giant AI Experiments: An Open Letter
The six-month pause letter Elon Musk signed before launching xAI in July 2023.
OpenAI sees $5 billion loss in 2024
OpenAI's projected 2024 loss of about $5 billion on about $3.7 billion of revenue.
Sam Altman says OpenAI is losing money on ChatGPT Pro
The $200-a-month tier that loses money.
Data centers accounted for nearly all US GDP growth in the first half of 2025
The 92% of US growth in the first half of 2025.
AI contributed "basically zero" to the US economy last year, according to Goldman Sachs
Why estimates differ: most AI equipment is imported, so it adds to Taiwanese and Korean output more than to American.
The AI capex boom is building twice as fast as the housing boom
The speed of the build-out, and the risk that it unwinds as fast.
03 · What is actually working
The Nobel Prize in Chemistry 2024
Demis Hassabis and John Jumper, for AlphaFold's protein-structure prediction.
Artificial intelligence-supported screen reading versus standard double reading in the MASAI trial
20% more cancers detected, no rise in false positives, 44% less reading workload.
Federal court allows collective action lawsuit over alleged AI hiring bias
The certification of the collective action against Workday.
Deloitte to partially refund Australian government for report with apparent AI-generated errors
The A$440,000 report with fabricated citations.
Klarna's AI bot is doing the work of 700 employees
Klarna's announcement, also referred to in chapters 04 and 05.
Klarna is hiring customer service agents after AI couldn't cut it
Klarna's reversal.
Goldman Sachs CIO on AI and developer productivity
GitHub Copilot for more than 9,000 developers, about 80% of the firm's engineers.
04 · Execution, the disciplined path
Regulation (EU) 2024/1689, the Artificial Intelligence Act
The obligations behind the compliance-mapping pillar.
ISO/IEC 42001:2023, AI management systems
A recognised structure for AI governance that auditors know.
The other widely used framework for AI risk.
Chapter 05 builds on the sources above and introduces none of its own.