AI boom

• 1. Introduction

  ›

  ChatGPT became the 4th-most visited website in the world, behind only Google, YouTube, and Facebook. It reached that position after growing to over 100 million users in just two months following its launch at the end of 2022, making it the fastest-growing software application on record. What does it mean when an AI chatbot outpaces every social network, every news outlet, every streaming platform? How did we arrive at a moment where a machine that writes, draws, codes, and composes music has become woven into the daily fabric of modern life? And what are the hidden costs, the contested claims, and the unresolved dangers that this rapid transformation carries with it?

• 2. The Long Road to Now

  ›

  Alan Turing proposed the idea of "Thinking Machines" in 1950, imagining computers capable of reasoning at the level of a human being. His Turing Test asked a simple but unsettling question: could an interrogator, given two sets of responses, reliably distinguish the human from the machine? Six years later, in 1956, John McCarthy gave the field its name. McCarthy, alongside Nathaniel Rochester, Marvin Minsky, and Claude Shannon, organized the Dartmouth conference, the gathering that formally constituted artificial intelligence as an academic discipline. McCarthy also created the programming language LISP in 1958, a tool that remained the dominant AI programming language in the United States for decades. He later founded the Stanford Artificial Intelligence Laboratory in 1962 and co-founded MIT's first AI lab, now known as the MIT Computer Science and Artificial Intelligence Laboratory. The earliest brush with conversational AI came in 1966, when Joseph Weizenbaum created ELIZA, designed as an experimental emotional tool and now recognized as the first chatbot. These were the seeds. What the early researchers could not have anticipated was how long the field would cycle between periods of excitement and contraction before arriving at the current era, which some call an "AI spring" to distinguish it from the "AI winters" that periodically cooled investment and interest.

• 3. What the Machines Can Do

  ›

  DeepMind's AlphaFold program scored more than 90 on CASP's Global Distance Test in 2020, a result that structural biologist and Nobel Prize winner Venki Ramakrishnan described as "a stunning advance on the protein folding problem." Predicting how a protein folds based on its amino acid sequence had resisted solution for decades; accurate prediction may accelerate drug discovery and improve understanding of disease. On the creative side, OpenAI released DALL-E in January 2021, giving users the ability to generate images from a plain text prompt. By 2022, DALL-E 2 and Midjourney had both launched. OpenAI's Sora, a text-to-video model, arrived in 2024 and was quickly taken up for advertising work, reducing production costs and increasing production speed. In audio, Google's DeepMind released WaveNet in 2016, a model that could generate raw speech and piano audio, identify individual speakers, and serve as a structural foundation for later voice generation tools. OpenAI released Jukebox in 2020, the first large-scale model capable of generating full songs across different genres and styles. By 2024, high-fidelity AI music tools had become publicly available. A 2026 study published in the journal Management Science found that less experienced software developers gained the largest productivity benefits from generative coding tools, with higher adoption rates and greater measured output gains.

• 4. The Business of Belief

  ›

  Nvidia's market capitalization rose above US$3.3 trillion as of the 19th of June 2024, making it the world's largest company by that measure at the time. The company's GPUs power the training and operation of generative AI models, meaning Nvidia's fortunes are directly tied to how much the industry expands. Nvidia later became the first company to reach US$4 trillion in market capitalization on the 9th of July 2025, and subsequently crossed US$5 trillion on the 29th of October of that same year. AI and generative AI investments grew from $18 billion in 2014 to $119 billion in 2021. Generative AI's share of total AI investment stood at around 30% in 2023. Google recognized the threat posed by ChatGPT to its core search business and responded by merging DeepMind and Google Brain, two internal research units that had previously operated separately. Stanford economist Erik Brynjolfsson has argued in a series of articles for a productivity boom driven by AI, while Northwestern economist Robert Gordon has taken a more pessimistic view; the two have registered a formal bet at Long Bets about productivity growth rates in the 2020s, to be settled at the end of the decade. Businesses report revenue increases of up to 16% linked to AI adoption, concentrated in manufacturing, risk management, and research and development. San Francisco's population increased in 2023 for the first time in years, with the AI boom cited as a factor. Despite this, research from MIT found that 95% of business AI projects were unprofitable in 2025, even as corporate investment in AI that year exceeded $60 billion.

• 5. Intellectual Property and Likeness

  ›

  Sarah Silverman's memoir The Bedwetter appeared in ChatGPT's outputs in detailed summaries and verbatim excerpts of paywalled journalism, raising suspicion about the sources embedded in the model's training data. Early generative chatbots, including GPT-1, drew on the BookCorpus; books remain, according to the source, the best training material for producing high-quality language models. Meta, OpenAI, and Nvidia have all faced lawsuits brought by artists, writers, journalists, and developers over the use of their work in training data. In protest at UK government consultations over how copyrighted music could legally be used in AI training, more than a thousand British musicians released an album containing no sounds at all, titled Is This What We Want? The conflict extended to voice and likeness. On the 19th of April 2024, Drake released a diss track called "Taylor Made Freestyle" featuring AI-generated vocals mimicking Tupac Shakur and Snoop Dogg. Shakur's estate threatened to sue, arguing it violated Shakur's personality rights. On the 20th of May 2024, actor Scarlett Johansson issued a statement accusing OpenAI of designing its "Sky" voice feature to closely resemble her voice and her portrayal of an AI assistant in the 2013 film Her, even though she had previously turned down the company's offer to provide that voice. OpenAI's unnamed voice actress, whose agent stated she recorded in her natural speaking voice, was drawn into a public dispute she had no part in creating. In late January 2024, deepfake images of Taylor Swift spread widely. The United States Congress introduced the DEFIANCE Act in March 2024, while Canada moved to introduce federal legislation targeting the non-consensual sharing of AI-generated explicit imagery.

• 6. Energy, Safety, and Extinction

  ›

  Google's greenhouse gas emissions rose by nearly 50% between 2019 and 2024, partly due to energy demands from AI data centers. Microsoft, Google, and Amazon have all invested in nuclear power to address their electricity needs. In September 2024, Microsoft signed a deal with Constellation Energy to purchase power from a reactor at Three Mile Island, a plant that had been shut down in 2019. That reactor is scheduled to reopen in 2028. It sits beside the unit responsible for the worst nuclear power accident in United States history, which occurred in 1979. The Center for AI Safety's researchers expect AI to improve the accessibility, success rate, scale, speed, stealth, and potency of cyberattacks, with potential for significant geopolitical disruption if the technology strengthens offensive capabilities more than defensive ones. Researchers have also raised concerns about the potential for future AI systems to assist in engineering highly lethal and contagious pathogens. A coalition of industry leaders and other signatories has endorsed the Statement on AI Risk, which argues that humanity could irreversibly lose control over a sufficiently advanced artificial general intelligence. A Pew Research report found that most Americans expressed concern about a lack of control over AI and over its potential negative effect on human creativity. A 2025 study by Nikolova and Angrisani found that people were especially distrustful of AI in personal relationships, while being more accepting of its role in medicine, such as developing new antibiotics. Economists Bell and Korinek argued in 2023 that the economic effects of AI could worsen inequality in ways that pose distinct threats to democratic governance, separate from the risks posed by AI-generated misinformation and propaganda.