GPT-3

9 results back to index

AI 2041 by Kai-Fu Lee, Chen Qiufan

3D printing, Abraham Maslow, active measures, airport security, Albert Einstein, AlphaGo, Any sufficiently advanced technology is indistinguishable from magic, artificial general intelligence, augmented reality, autonomous vehicles, basic income, bitcoin, blockchain, blue-collar work, Cambridge Analytica, carbon footprint, Charles Babbage, computer vision, contact tracing, coronavirus, corporate governance, corporate social responsibility, COVID-19, CRISPR, cryptocurrency, DALL-E, data science, deep learning, deepfake, DeepMind, delayed gratification, dematerialisation, digital map, digital rights, digital twin, Elon Musk, fake news, fault tolerance, future of work, Future Shock, game design, general purpose technology, global pandemic, Google Glasses, Google X / Alphabet X, GPT-3, happiness index / gross national happiness, hedonic treadmill, hiring and firing, Hyperloop, information security, Internet of things, iterative process, job automation, language acquisition, low earth orbit, Lyft, Maslow's hierarchy, mass immigration, mirror neurons, money: store of value / unit of account / medium of exchange, mutually assured destruction, natural language processing, Neil Armstrong, Nelson Mandela, OpenAI, optical character recognition, pattern recognition, plutocrats, post scarcity, profit motive, QR code, quantitative easing, Richard Feynman, ride hailing / ride sharing, robotic process automation, Satoshi Nakamoto, self-driving car, seminal paper, Silicon Valley, smart cities, smart contracts, smart transportation, Snapchat, social distancing, speech recognition, Stephen Hawking, synthetic biology, telemarketer, Tesla Model S, The future is already here, trolley problem, Turing test, uber lyft, universal basic income, warehouse automation, warehouse robotics, zero-sum game

And this 500,000-lifetimes figure is increasing by ten times every year, adding capabilities at an unbelievable exponential pace. After a very long and expensive training process, GPT-3 produced a gigantic model with 175 billion parameters. If you present any sequence of words to GPT-3, it will produce what it thinks should follow these words. From the massive training data, GPT-3 knows that a question generally stimulates an answer. For example, if you told GPT-3: “A stove is heavier than a cat. An ocean is heavier than a dust particle. Which is heavier, a toaster or a pencil?” GPT-3 will correctly answer “a toaster.” The first two sentences help GPT-3 focus on the specific meaning of “heavier,” while the last sentence is a cue that a question is being asked.

…

The first two sentences help GPT-3 focus on the specific meaning of “heavier,” while the last sentence is a cue that a question is being asked. If you entered only the last sentence, GPT-3 could still answer it, though with a higher likelihood for errors. GPT-3 differs dramatically from domain-specific NLP. Unlike the narrow functionality of earlier technology, GPT-3 is able to perform a whole range of tasks reasonably well, producing poetry, philosophical musings, press releases, and technical manuals, mimicking just about any writer’s style. For example, a reporter asked GPT-3 to write a Dr. Seuss–style poem about Elon Musk: But then, in his haste, he got into a fight.

…

With enough natural data and sufficient processing power, the system can learn on its own to detect arrival and departure times, and a great deal more. After Google’s transformer work, a more well-known extension called GPT-3 (GPT stands for “generative pre-trained transformers”) was released in 2020 by OpenAI, a research laboratory founded by Elon Musk and others. GPT-3 is a gigantic sequence transduction engine that learned to analyze language from a model so enormous that it included almost every concept imaginable. Leveraging one of the most powerful supercomputers in the world, GPT-3 was trained on more than 45 terabytes of text, which would take 500,000 lifetimes for a human to read. And this 500,000-lifetimes figure is increasing by ten times every year, adding capabilities at an unbelievable exponential pace.

System Error by Rob Reich

"Friedman doctrine" OR "shareholder theory", "World Economic Forum" Davos, 2021 United States Capitol attack, A Declaration of the Independence of Cyberspace, Aaron Swartz, AI winter, Airbnb, airport security, Alan Greenspan, Albert Einstein, algorithmic bias, AlphaGo, AltaVista, artificial general intelligence, Automated Insights, autonomous vehicles, basic income, Ben Horowitz, Berlin Wall, Bernie Madoff, Big Tech, bitcoin, Blitzscaling, Cambridge Analytica, Cass Sunstein, clean water, cloud computing, computer vision, contact tracing, contact tracing app, coronavirus, corporate governance, COVID-19, creative destruction, CRISPR, crowdsourcing, data is the new oil, data science, decentralized internet, deep learning, deepfake, DeepMind, deplatforming, digital rights, disinformation, disruptive innovation, Donald Knuth, Donald Trump, driverless car, dual-use technology, Edward Snowden, Elon Musk, en.wikipedia.org, end-to-end encryption, Fairchild Semiconductor, fake news, Fall of the Berlin Wall, Filter Bubble, financial engineering, financial innovation, fulfillment center, future of work, gentrification, Geoffrey Hinton, George Floyd, gig economy, Goodhart's law, GPT-3, Hacker News, hockey-stick growth, income inequality, independent contractor, informal economy, information security, Jaron Lanier, Jeff Bezos, Jim Simons, jimmy wales, job automation, John Maynard Keynes: Economic Possibilities for our Grandchildren, John Maynard Keynes: technological unemployment, John Perry Barlow, Lean Startup, linear programming, Lyft, Marc Andreessen, Mark Zuckerberg, meta-analysis, minimum wage unemployment, Monkeys Reject Unequal Pay, move fast and break things, Myron Scholes, Network effects, Nick Bostrom, Northpointe / Correctional Offender Management Profiling for Alternative Sanctions, NP-complete, Oculus Rift, OpenAI, Panopticon Jeremy Bentham, Parler "social media", pattern recognition, personalized medicine, Peter Thiel, Philippa Foot, premature optimization, profit motive, quantitative hedge fund, race to the bottom, randomized controlled trial, recommendation engine, Renaissance Technologies, Richard Thaler, ride hailing / ride sharing, Ronald Reagan, Sam Altman, Sand Hill Road, scientific management, self-driving car, shareholder value, Sheryl Sandberg, Shoshana Zuboff, side project, Silicon Valley, Snapchat, social distancing, Social Responsibility of Business Is to Increase Its Profits, software is eating the world, spectrum auction, speech recognition, stem cell, Steve Jobs, Steven Levy, strong AI, superintelligent machines, surveillance capitalism, Susan Wojcicki, tech billionaire, tech worker, techlash, technoutopianism, Telecommunications Act of 1996, telemarketer, The Future of Employment, TikTok, Tim Cook: Apple, traveling salesman, Triangle Shirtwaist Factory, trolley problem, Turing test, two-sided market, Uber and Lyft, uber lyft, ultimatum game, union organizing, universal basic income, washing machines reduced drudgery, Watson beat the top human players on Jeopardy!, When a measure becomes a target, winner-take-all economy, Y Combinator, you are the product

the training data: Stephen Ornes, “Explainer: Understanding the Size of Data,” Science News for Students, December 13, 2013, https://www.sciencenewsforstudents.org/article/explainer-understanding-size-data. “Kanye West Exclusive”: Arram Sabeti, “GPT-3,” Arram Sabeti (blog), July 9, 2020, https://arr.am/2020/07/09/gpt-3-an-ai-thats-eerily-good-at-writing-almost-anything/. “Why deep learning will never”: Gwern Branwen, “GPT-3 Creative Fiction,” gwern.net, June 19, 2020, https://www.gwern.net/GPT-3#why-deep-learning-will-never-truly-x; Kelsey Piper, “GPT-3, Explained: This New Language AI Is Uncanny, Funny—and a Big Deal,” Vox, August 13, 2020, https://www.vox.com/future-perfect/21355768/gpt-3-ai-openai-turing-test-language. trust in technology companies is declining: Carroll Doherty and Jocelyn Kiley, “Americans Have Become Much Less Positive About Tech Companies’ Impact on the U.S.,” Pew Research Center, July 29, 2019, https://www.pewresearch.org/fact-tank/2019/07/29/americans-have-become-much-less-positive-about-tech-companies-impact-on-the-u-s/; Ina Fried, “40% of Americans Believe Artificial Intelligence Needs More Regulation,” Axios, https://www.axios.com/big-tech-industry-global-trust-9b7c6c3c-98f1-4e80-8275-cf52446b1515.html.

…

About a year later, the OpenAI team announced GPT-3, an exponentially more powerful next-generation model with more than one hundred times as many parameters as the largest model used in GPT-2. The deep learning ninety-six-layer neural network that forms the basis of GPT-3 was trained on a huge volume of text taken from the internet in addition to an enormous library of books and the entirety of Wikipedia. To give a sense of scale, the training data for GPT-3 is nearly 45 terabytes in size, or more than four times the estimated size of all the printed material in the Library of Congress in 2000. GPT-3 represents an important frontier in AI research.

…

It was generated by GPT-3 in response to the prompt “Why deep learning will never truly X.” It can also craft Harry Potter stories in the style of Ernest Hemingway, invent plausible conversations between famous people in history who never met, summarize movies with emojis, write poetry, and much more. The reason we know about these capabilities is that OpenAI released the GPT-3 model to interested parties, albeit through an application process in which OpenAI controls access. Those granted access began playing with it and posting their findings. OpenAI announced its intention to offer GPT-3 as a revenue-generating commercial product in limited contexts.

Rule of the Robots: How Artificial Intelligence Will Transform Everything by Martin Ford

AI winter, Airbnb, algorithmic bias, algorithmic trading, Alignment Problem, AlphaGo, Amazon Mechanical Turk, Amazon Web Services, artificial general intelligence, Automated Insights, autonomous vehicles, backpropagation, basic income, Big Tech, big-box store, call centre, carbon footprint, Chris Urmson, Claude Shannon: information theory, clean water, cloud computing, commoditize, computer age, computer vision, Computing Machinery and Intelligence, coronavirus, correlation does not imply causation, COVID-19, crowdsourcing, data is the new oil, data science, deep learning, deepfake, DeepMind, Demis Hassabis, deskilling, disruptive innovation, Donald Trump, Elon Musk, factory automation, fake news, fulfillment center, full employment, future of work, general purpose technology, Geoffrey Hinton, George Floyd, gig economy, Gini coefficient, global pandemic, Googley, GPT-3, high-speed rail, hype cycle, ImageNet competition, income inequality, independent contractor, industrial robot, informal economy, information retrieval, Intergovernmental Panel on Climate Change (IPCC), Internet of things, Jeff Bezos, job automation, John Markoff, Kiva Systems, knowledge worker, labor-force participation, Law of Accelerating Returns, license plate recognition, low interest rates, low-wage service sector, Lyft, machine readable, machine translation, Mark Zuckerberg, Mitch Kapor, natural language processing, Nick Bostrom, Northpointe / Correctional Offender Management Profiling for Alternative Sanctions, Ocado, OpenAI, opioid epidemic / opioid crisis, passive income, pattern recognition, Peter Thiel, Phillips curve, post scarcity, public intellectual, Ray Kurzweil, recommendation engine, remote working, RFID, ride hailing / ride sharing, Robert Gordon, Rodney Brooks, Rubik’s Cube, Sam Altman, self-driving car, Silicon Valley, Silicon Valley startup, social distancing, SoftBank, South of Market, San Francisco, special economic zone, speech recognition, stealth mode startup, Stephen Hawking, superintelligent machines, TED Talk, The Future of Employment, The Rise and Fall of American Growth, the scientific method, Turing machine, Turing test, Tyler Cowen, Tyler Cowen: Great Stagnation, Uber and Lyft, uber lyft, universal basic income, very high income, warehouse automation, warehouse robotics, Watson beat the top human players on Jeopardy!, WikiLeaks, women in the workforce, Y Combinator

It soon became clear, however, that many of the most impressive examples had been cherry-picked from multiple trials, and that GPT-3, like its predecessor, often produced coherently written nonsense. Both of OpenAI’s GPT systems are at their core powerful prediction engines. Given a sequence of words, they excel at predicting what the next word should be. GPT-3 takes this capability to an unprecedented level, and because the massive trove of text the system was trained on encapsulates real knowledge, the system does often produce very useful output. There is no consistency, however, and GPT-3 often generates rubbish and struggles with tasks that would be simple for any human.39 Compared to its predecessor, GPT-3 can certainly write a far more compelling story about unicorns.

…

The system does not know what a unicorn is, or that a “four-horned” variety would contradict that meaning. GPT-2 suffers from the same fundamental limitation that David Ferrucci’s team at Elemental Cognition and Ray Kurzweil at Google are trying to address. In May 2020, OpenAI released GPT-3, a vastly more powerful system. While GPT-2’s neural network included about 1.5 billion weights that were optimized as the network was trained, GPT-3 increased that number more than a hundredfold to 175 billion. GPT-3’s neural network was trained on about half a terabyte of text, an amount so vast that the entire English version of Wikipedia—roughly six million articles—constitutes only about 0.6 percent of the total.

…

GPT-3’s neural network was trained on about half a terabyte of text, an amount so vast that the entire English version of Wikipedia—roughly six million articles—constitutes only about 0.6 percent of the total. OpenAI offered early access to a select group of AI researchers and journalists and announced plans to eventually turn the new system into its first commercial product. Over the next few weeks, as people began to experiment with GPT-3, social media exploded with astonishment at the power of the new system. Given the proper prompts, GPT-3 could write convincing articles or poems in the style of long-dead authors. It could even generate faux conversations between historical or fictional figures. A college student used the system to generate all the posts for a self-help blog that rose to the top of the charts.38 All this quickly led to speculation that the system represented a critical breakthrough on the path to human-level machine intelligence.

Four Battlegrounds by Paul Scharre

2021 United States Capitol attack, 3D printing, active measures, activist lawyer, AI winter, AlphaGo, amateurs talk tactics, professionals talk logistics, artificial general intelligence, ASML, augmented reality, Automated Insights, autonomous vehicles, barriers to entry, Berlin Wall, Big Tech, bitcoin, Black Lives Matter, Boeing 737 MAX, Boris Johnson, Brexit referendum, business continuity plan, business process, carbon footprint, chief data officer, Citizen Lab, clean water, cloud computing, commoditize, computer vision, coronavirus, COVID-19, crisis actor, crowdsourcing, DALL-E, data is not the new oil, data is the new oil, data science, deep learning, deepfake, DeepMind, Demis Hassabis, Deng Xiaoping, digital map, digital rights, disinformation, Donald Trump, drone strike, dual-use technology, Elon Musk, en.wikipedia.org, endowment effect, fake news, Francis Fukuyama: the end of history, future of journalism, future of work, game design, general purpose technology, Geoffrey Hinton, geopolitical risk, George Floyd, global supply chain, GPT-3, Great Leap Forward, hive mind, hustle culture, ImageNet competition, immigration reform, income per capita, interchangeable parts, Internet Archive, Internet of things, iterative process, Jeff Bezos, job automation, Kevin Kelly, Kevin Roose, large language model, lockdown, Mark Zuckerberg, military-industrial complex, move fast and break things, Nate Silver, natural language processing, new economy, Nick Bostrom, one-China policy, Open Library, OpenAI, PalmPilot, Parler "social media", pattern recognition, phenotype, post-truth, purchasing power parity, QAnon, QR code, race to the bottom, RAND corporation, recommendation engine, reshoring, ride hailing / ride sharing, robotic process automation, Rodney Brooks, Rubik’s Cube, self-driving car, Shoshana Zuboff, side project, Silicon Valley, slashdot, smart cities, smart meter, Snapchat, social software, sorting algorithm, South China Sea, sparse data, speech recognition, Steve Bannon, Steven Levy, Stuxnet, supply-chain attack, surveillance capitalism, systems thinking, tech worker, techlash, telemarketer, The Brussels Effect, The Signal and the Noise by Nate Silver, TikTok, trade route, TSMC

The hype surrounding deepfakes may have outpaced their reality today, but long-term trends in data and computing power will enable more powerful and accessible machine learning tools to create convincing synthetic media. The AI text generator GPT-2, whose staged release caused such a stir in 2019, was eclipsed only fifteen months later by GPT-3, a 175 billion parameter model that was over ten times larger than GPT-2. GPT-3’s text is shockingly convincing. Renée DiResta, technical research manager at the Stanford Internet Observatory, prompted GPT-3 to weigh in on the implications of synthetic media. GPT-3’s response: AI - generated content will continue to become more sophisticated, and it will be increasingly difficult to differentiate it from the content that is created by humans.

…

The amount of compute used for training cutting-edge machine learning research projects increased ten billionfold from 2010 to 2022 and is doubling roughly every six months. Compute for training the largest models, like GPT-3 and PaLM, has been doubling at a slightly slower rate, approximately every ten months. This is an incredible explosion of compute, yet there are likely limits to how long compute usage can grow at this pace. Compute-intensive research is expensive, and even the most deep-pocketed actors have limits to their resources. Independent estimates put the cost to train advanced machine learning models such as GPT-3 on the order of millions of dollars per research project for some of the largest models. These costs already put compute-intensive research out of reach for some actors, such as universities.

…

(Coco Feng, “US-China tech war: Beijing-Funded AI Researchers Surpass Google and OpenAI with New Language Processing model,” South China Morning Post, June 2, 2021, https://www.scmp.com/tech/tech-war/article/3135764/us-china-tech-war-beijing-funded-ai-researchers-surpass-google-and; Alberto Romero, “GPT-3 Scared You? Meet Wu Dao 2.0: A Monster of 1.75 Trillion Parameters,” Towards Data Science, June 5, 2021, https://towardsdatascience.com/gpt-3-scared-you-meet-wu-dao-2-0-a-monster-of-1-75-trillion-parameters-832cd83db484.) In April 2022, researchers from several labs, including Tsinghua University, Alibaba, and the Beijing Academy of Artificial Intelligence, announced a framework for scaling up training to 14.5 trillion parameter models, with the long-term intent of training “brain scale” models of over 100 trillion parameters.

Being You: A New Science of Consciousness by Anil Seth

AlphaGo, artificial general intelligence, augmented reality, backpropagation, carbon-based life, Claude Shannon: information theory, computer age, computer vision, Computing Machinery and Intelligence, coronavirus, correlation does not imply causation, CRISPR, cryptocurrency, deep learning, deepfake, DeepMind, Drosophila, en.wikipedia.org, Filter Bubble, GPT-3, GPT-4, John Markoff, longitudinal study, Louis Pasteur, mirror neurons, Neil Armstrong, Nick Bostrom, Norbert Wiener, OpenAI, paperclip maximiser, pattern recognition, Paul Graham, Pierre-Simon Laplace, planetary scale, Plato's cave, precautionary principle, Ray Kurzweil, self-driving car, speech recognition, stem cell, systems thinking, technological singularity, TED Talk, telepresence, the scientific method, theory of mind, Thomas Bayes, TikTok, Turing test

As AI continues to improve, the Turing test may soon be passed without such artificially low standards. In May 2020, the research lab OpenAI released GPT-3 – a vast artificial neural network trained on examples of natural language drawn from a large swathe of the internet. As well as engaging in chatbot-variety dialogue, GPT-3 can generate substantial passages of text in many different styles when prompted with a few initial words or lines. Although it does not understand what it produces, the fluency and sophistication of GPT-3’s output is surprising and, for some, even frightening. In one example, published in the Guardian, it delivered a five-hundred-word essay about why humans should not be afraid of AI – ranging across topics from the psychology of human violence to the industrial revolution, and including the disconcerting line: ‘AI should not waste time trying to understand the viewpoints of people who distrust artificial intelligence for a living.’

…

These networks are trained using an unsupervised deep learning approach essentially to ‘predict the next word’ given a previous word or text snippet. GPT-3 has an astonishing 175 billion parameters and was trained on some 45 terabytes of text data. See https://openai.com/blog/openai-api/ and for technical details: https://arxiv.org/abs/2005.14165. it does not understand: Of course this depends on what is meant by ‘understanding’. Some might say that human ‘understanding’ is no different in kind from the sort of ‘understanding’ displayed by GPT-3. The cognitive scientist Gary Marcus argues against this position, and I agree with him. See www.technologyreview.com/2020/08/22/1007539/gpt3-openai-language-generator-artificial-intelligence-ai-opinion/.

…

In one example, published in the Guardian, it delivered a five-hundred-word essay about why humans should not be afraid of AI – ranging across topics from the psychology of human violence to the industrial revolution, and including the disconcerting line: ‘AI should not waste time trying to understand the viewpoints of people who distrust artificial intelligence for a living.’ Despite its sophistication, I am pretty sure that GPT-3 can still be caught out by any reasonably sophisticated human interlocutor. This may not be true for GPT-4, or GPT-10. But even if a future GPT-like system repeatedly aces the Turing test, it would be exhibiting only a very narrow form of (simulated) intelligence – disembodied linguistic exchange – rather than the fully embodied ‘doing the right thing at the right time’ natural intelligence that we see in humans and in many other animals – as well as in my hypothetical silicon beast machine.

The Coming Wave: Technology, Power, and the Twenty-First Century's Greatest Dilemma by Mustafa Suleyman

"World Economic Forum" Davos, 23andMe, 3D printing, active measures, Ada Lovelace, additive manufacturing, agricultural Revolution, AI winter, air gap, Airbnb, Alan Greenspan, algorithmic bias, Alignment Problem, AlphaGo, Alvin Toffler, Amazon Web Services, Anthropocene, artificial general intelligence, Asilomar, Asilomar Conference on Recombinant DNA, ASML, autonomous vehicles, backpropagation, barriers to entry, basic income, benefit corporation, Big Tech, biodiversity loss, bioinformatics, Bletchley Park, Blitzscaling, Boston Dynamics, business process, business process outsourcing, call centre, Capital in the Twenty-First Century by Thomas Piketty, ChatGPT, choice architecture, circular economy, classic study, clean tech, cloud computing, commoditize, computer vision, coronavirus, corporate governance, correlation does not imply causation, COVID-19, creative destruction, CRISPR, critical race theory, crowdsourcing, cryptocurrency, cuban missile crisis, data science, decarbonisation, deep learning, deepfake, DeepMind, deindustrialization, dematerialisation, Demis Hassabis, disinformation, drone strike, drop ship, dual-use technology, Easter island, Edward Snowden, effective altruism, energy transition, epigenetics, Erik Brynjolfsson, Ernest Rutherford, Extinction Rebellion, facts on the ground, failed state, Fairchild Semiconductor, fear of failure, flying shuttle, Ford Model T, future of work, general purpose technology, Geoffrey Hinton, global pandemic, GPT-3, GPT-4, hallucination problem, hive mind, hype cycle, Intergovernmental Panel on Climate Change (IPCC), Internet Archive, Internet of things, invention of the wheel, job automation, John Maynard Keynes: technological unemployment, John von Neumann, Joi Ito, Joseph Schumpeter, Kickstarter, lab leak, large language model, Law of Accelerating Returns, Lewis Mumford, license plate recognition, lockdown, machine readable, Marc Andreessen, meta-analysis, microcredit, move 37, Mustafa Suleyman, mutually assured destruction, new economy, Nick Bostrom, Nikolai Kondratiev, off grid, OpenAI, paperclip maximiser, personalized medicine, Peter Thiel, planetary scale, plutocrats, precautionary principle, profit motive, prompt engineering, QAnon, quantum entanglement, ransomware, Ray Kurzweil, Recombinant DNA, Richard Feynman, Robert Gordon, Ronald Reagan, Sam Altman, Sand Hill Road, satellite internet, Silicon Valley, smart cities, South China Sea, space junk, SpaceX Starlink, stealth mode startup, stem cell, Stephen Fry, Steven Levy, strong AI, synthetic biology, tacit knowledge, tail risk, techlash, techno-determinism, technoutopianism, Ted Kaczynski, the long tail, The Rise and Fall of American Growth, Thomas Malthus, TikTok, TSMC, Turing test, Tyler Cowen, Tyler Cowen: Great Stagnation, universal basic income, uranium enrichment, warehouse robotics, William MacAskill, working-age population, world market for maybe five computers, zero day

., “Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity,” Journal of Machine Learning Research, June 16, 2022, arxiv.org/​abs/​2101.03961. GO TO NOTE REFERENCE IN TEXT Or look at DeepMind’s Chinchilla Alberto Romero, “A New AI Trend: Chinchilla (70B) Greatly Outperforms GPT-3 (175B) and Gopher (280B),” Towards Data Science, April 11, 2022, towardsdatascience.com/​a-new-ai-trend-chinchilla-70b-greatly-outperforms-gpt-3-175b-and-gopher-280b-408b9b4510. GO TO NOTE REFERENCE IN TEXT At the other end of the spectrum See github.com/​karpathy/​nanoGPT for more details. GO TO NOTE REFERENCE IN TEXT Meta has open-sourced Susan Zhang et al., “Democratizing Access to Large-Scale Language Models with OPT-175B,” Meta AI, May 3, 2022, ai.facebook.com/​blog/​democratizing-access-to-large-scale-language-models-with-opt-175b.

…

(GPT stands for generative pre-trained transformer.) It was, at the time, an enormous model. With 1.5 billion parameters (the number of parameters is a core measure of an AI system’s scale and complexity), GPT-2 was trained on 8 million pages of web text. But it wasn’t until the summer of 2020, when OpenAI released GPT-3, that people started to truly grasp the magnitude of what was happening. With a whopping 175 billion parameters it was, at the time, the largest neural network ever constructed, more than a hundred times larger than its predecessor of just a year earlier. Impressive, yes, but that scale is now routine, and the cost of training an equivalent model has fallen tenfold over the last two years.

…

MORE WITH LESS, AGAIN When a new technology starts working, it always becomes dramatically more efficient. AI is no different. Google’s Switch Transformer, for example, has 1.6 trillion parameters. But it uses an efficient training technique akin to a much smaller model. At Inflection AI we can reach GPT-3-level language model performance with a system just one twenty-fifth the size. We have a model that beats Google’s 540 billion parameter PaLM on all the main academic benchmarks, but is six times smaller. Or look at DeepMind’s Chinchilla model, competitive with the very best large models, which has four times fewer parameters than its Gopher model, but instead uses more training data.

A Hacker's Mind: How the Powerful Bend Society's Rules, and How to Bend Them Back by Bruce Schneier

4chan, Airbnb, airport security, algorithmic trading, Alignment Problem, AlphaGo, Automated Insights, banking crisis, Big Tech, bitcoin, blockchain, Boeing 737 MAX, Brian Krebs, Capital in the Twenty-First Century by Thomas Piketty, cloud computing, computerized trading, coronavirus, corporate personhood, COVID-19, cryptocurrency, dark pattern, deepfake, defense in depth, disinformation, Donald Trump, Double Irish / Dutch Sandwich, driverless car, Edward Thorp, Elon Musk, fake news, financial innovation, Financial Instability Hypothesis, first-past-the-post, Flash crash, full employment, gig economy, global pandemic, Goodhart's law, GPT-3, Greensill Capital, high net worth, Hyman Minsky, income inequality, independent contractor, index fund, information security, intangible asset, Internet of things, Isaac Newton, Jeff Bezos, job automation, late capitalism, lockdown, Lyft, Mark Zuckerberg, money market fund, moral hazard, move fast and break things, Nate Silver, offshore financial centre, OpenAI, payday loans, Peter Thiel, precautionary principle, Ralph Nader, recommendation engine, ride hailing / ride sharing, self-driving car, sentiment analysis, Skype, smart cities, SoftBank, supply chain finance, supply-chain attack, surveillance capitalism, systems thinking, TaskRabbit, technological determinism, TED Talk, The Wealth of Nations by Adam Smith, theory of mind, TikTok, too big to fail, Turing test, Uber and Lyft, uber lyft, ubercab, UNCLOS, union organizing, web application, WeWork, When a measure becomes a target, WikiLeaks, zero day

For years, AI programs have composed news stories about sports and finance for real news organizations like the Associated Press. The constrained nature of much reporting on those topics has made them easier to adapt to AI. AI is now being used to write more general stories. Modern text-creation systems like Open AI’s GPT-3 can be fed facts and write true stories, but they can just as easily be fed untruths and write fake news. It doesn’t take much imagination to see how AI will degrade political discourse. Already, AI-driven personas can write personalized letters to newspapers and elected officials, leave intelligible comments on news sites and message boards, and intelligently debate politics on social media.

…

Persona bots have histories, personalities, and communication styles. They don’t constantly spew propaganda. They hang out in various interest groups: gardening, knitting, model railroading, whatever. They act as normal members of those communities, posting and commenting and discussing. Systems like GPT-3 will make it easy for those AIs to mine previous conversations and related Internet content and to appear knowledgeable. Then, once in a while, the AI might post something relevant to a political issue, maybe an article about a healthcare worker having an allergic reaction to the COVID-19 vaccine, with worried commentary.

…

The AI needs some sort of feedback on how well it is doing so that it can improve its performance. Sometimes this is a trivial matter. For a game like Go, it’s easy. The rules, objective, and feedback—did you win or lose?—are all precisely specified, and there’s nothing outside of those things to muddy the waters. The GPT-3 AI can write coherent essays because its “world” is just text. This is why most of the current examples of goal and reward hacking come from simulated environments. Those are artificial and constrained, with all of the rules specified to the AI. What matters is the amount of ambiguity in a system.

Power and Progress: Our Thousand-Year Struggle Over Technology and Prosperity by Daron Acemoglu, Simon Johnson

"Friedman doctrine" OR "shareholder theory", "World Economic Forum" Davos, 4chan, agricultural Revolution, AI winter, Airbnb, airline deregulation, algorithmic bias, algorithmic management, Alignment Problem, AlphaGo, An Inconvenient Truth, artificial general intelligence, augmented reality, basic income, Bellingcat, Bernie Sanders, Big Tech, Bletchley Park, blue-collar work, British Empire, carbon footprint, carbon tax, carried interest, centre right, Charles Babbage, ChatGPT, Clayton Christensen, clean water, cloud computing, collapse of Lehman Brothers, collective bargaining, computer age, Computer Lib, Computing Machinery and Intelligence, conceptual framework, contact tracing, Corn Laws, Cornelius Vanderbilt, coronavirus, corporate social responsibility, correlation does not imply causation, cotton gin, COVID-19, creative destruction, declining real wages, deep learning, DeepMind, deindustrialization, Demis Hassabis, Deng Xiaoping, deskilling, discovery of the americas, disinformation, Donald Trump, Douglas Engelbart, Douglas Engelbart, Edward Snowden, Elon Musk, en.wikipedia.org, energy transition, Erik Brynjolfsson, European colonialism, everywhere but in the productivity statistics, factory automation, facts on the ground, fake news, Filter Bubble, financial innovation, Ford Model T, Ford paid five dollars a day, fulfillment center, full employment, future of work, gender pay gap, general purpose technology, Geoffrey Hinton, global supply chain, Gordon Gekko, GPT-3, Grace Hopper, Hacker Ethic, Ida Tarbell, illegal immigration, income inequality, indoor plumbing, industrial robot, interchangeable parts, invisible hand, Isaac Newton, Jacques de Vaucanson, James Watt: steam engine, Jaron Lanier, Jeff Bezos, job automation, Johannes Kepler, John Markoff, John Maynard Keynes: Economic Possibilities for our Grandchildren, John Maynard Keynes: technological unemployment, Joseph-Marie Jacquard, Kenneth Arrow, Kevin Roose, Kickstarter, knowledge economy, labor-force participation, land reform, land tenure, Les Trente Glorieuses, low skilled workers, low-wage service sector, M-Pesa, manufacturing employment, Marc Andreessen, Mark Zuckerberg, megacity, mobile money, Mother of all demos, move fast and break things, natural language processing, Neolithic agricultural revolution, Norbert Wiener, NSO Group, offshore financial centre, OpenAI, PageRank, Panopticon Jeremy Bentham, paperclip maximiser, pattern recognition, Paul Graham, Peter Thiel, Productivity paradox, profit maximization, profit motive, QAnon, Ralph Nader, Ray Kurzweil, recommendation engine, ride hailing / ride sharing, Robert Bork, Robert Gordon, Robert Solow, robotic process automation, Ronald Reagan, scientific management, Second Machine Age, self-driving car, seminal paper, shareholder value, Sheryl Sandberg, Shoshana Zuboff, Silicon Valley, social intelligence, Social Responsibility of Business Is to Increase Its Profits, social web, South Sea Bubble, speech recognition, spice trade, statistical model, stem cell, Steve Jobs, Steve Wozniak, strikebreaker, subscription business, Suez canal 1869, Suez crisis 1956, supply-chain management, surveillance capitalism, tacit knowledge, tech billionaire, technoutopianism, Ted Nelson, TED Talk, The Future of Employment, The Rise and Fall of American Growth, The Structural Transformation of the Public Sphere, The Wealth of Nations by Adam Smith, theory of mind, Thomas Malthus, too big to fail, total factor productivity, trade route, transatlantic slave trade, trickle-down economics, Turing machine, Turing test, Twitter Arab Spring, Two Sigma, Tyler Cowen, Tyler Cowen: Great Stagnation, union organizing, universal basic income, Unsafe at Any Speed, Upton Sinclair, upwardly mobile, W. E. B. Du Bois, War on Poverty, WikiLeaks, wikimedia commons, working poor, working-age population

Some experts define artificial intelligence as machines or algorithms demonstrating “intelligent behavior” or “high-level capabilities,” although what these are is often open to debate. Others provide definitions motivated by programs such as GPT-3, equating intelligent machines with those that have goals, observe their environment, obtain other inputs, and attempt to achieve their objectives. For example, GPT-3 receives distinct goals in different applications and tries to accomplish them as successfully as possible. Whatever the exact definition of modern machine intelligence, it is clear that new digital algorithms are being applied widely to every domain of our lives.

…

Although AlphaZero’s chess moves within the rules of the game are impressive, they do not involve the type of creativity that humans regularly engage in—such as drawing analogies across unstructured, disparate environments and coming up with solutions to new and varied problems. Even GPT-3, though more versatile and impressive than AlphaZero, shows the same limitations. It cannot perform tasks beyond those for which it has been pretrained and shows no judgment, so conflicting or unusual instructions can stump it. Worse, this technology has no element of the social or situational intelligence of humans. GPT-3 cannot reason about the context in which the tasks it is performing are situated and draw on causal relationships that exist between actions and effects.

…

From the Field of AI Dreams People are right to be excited about advances in digital technologies. New machine capabilities can massively expand the things we do and can transform many aspects of our lives for the better. And there have also been tremendous advances. For example, the Generative Pre-trained Transformer 3 (GPT-3), released in 2020 by OpenAI, and ChatGPT released in 2022 by the same company, are natural-language processing systems with remarkable capabilities. Already trained and optimized on massive amounts of text data from the internet, these programs can generate almost human-like articles, including poetry; communicate in typical human language; and, most impressively, turn natural-language instructions into computer code.

Shape: The Hidden Geometry of Information, Biology, Strategy, Democracy, and Everything Else by Jordan Ellenberg

Albert Einstein, AlphaGo, Andrew Wiles, autonomous vehicles, British Empire, Brownian motion, Charles Babbage, Claude Shannon: information theory, computer age, coronavirus, COVID-19, deep learning, DeepMind, Donald Knuth, Donald Trump, double entry bookkeeping, East Village, Edmond Halley, Edward Jenner, Elliott wave, Erdős number, facts on the ground, Fellow of the Royal Society, Geoffrey Hinton, germ theory of disease, global pandemic, government statistician, GPT-3, greed is good, Henri Poincaré, index card, index fund, Isaac Newton, Johannes Kepler, John Conway, John Nash: game theory, John Snow's cholera map, Louis Bachelier, machine translation, Mercator projection, Mercator projection distort size, especially Greenland and Africa, Milgram experiment, multi-armed bandit, Nate Silver, OpenAI, Paul Erdős, pets.com, pez dispenser, probability theory / Blaise Pascal / Pierre de Fermat, Ralph Nelson Elliott, random walk, Rubik’s Cube, self-driving car, side hustle, Snapchat, social distancing, social graph, transcontinental railway, urban renewal

And yet, modern-day Markov chains can produce something remarkably like human language. An algorithm like Open AI’s GPT-3 is the spiritual descendant of Shannon’s text machine, only much bigger. The input, instead of being three letters, is a chunk of text hundreds of words long, but the principle is the same: given the passage of text most recently produced, what is the probability that the next word is “the,” or “geometry,” or “graupel”? You might think that this would be easy. You could take the first five sentences from your book and run them through GPT-3, and you’d get back a list of probabilities for every possible combination of words in those sentences.

…

You could take the first five sentences from your book and run them through GPT-3, and you’d get back a list of probabilities for every possible combination of words in those sentences. Wait, why would you think it would be easy? You wouldn’t, actually. That paragraph above is GPT-3’s attempt to continue on from the three paragraphs before it. I picked the most sensible output out of about ten tries. But all the outputs somehow do sound like they come from the book you’re reading, which, let me tell you, is somewhat unsettling for the human being writing the book, even when the sentences make no literal sense at all, as in this GPT-3 output: If you’re familiar with the concept of Bayes’ theorem, then this should be easy for you. If there’s a 50% chance that the next word will be “the” and a 50% chance that it’ll be “geometry,” then the probability that the next word is either “the geometry” or “graupel” is (50/50)2 = 0.

…

You can see that this might be a good idea for mountain climbing (though not always—we’ll come back to this), but what does it have to do with machine learning? Let’s say I’m not a mountaineer after all, but a computer trying to learn something. It might be one of the machines we’ve already encountered, like AlphaGo, the machine that learns to play Go better than a master, or GPT-3, the machine that produces long strings of discomfitingly plausible English text. But to start with, let’s stick with the classics, and suppose I’m a computer trying to learn what a cat is. How am I supposed to do that? The same way a baby does, more or less. The baby lives in a world where every so often some large person points at something in their visual field and says “Cat!”

Futureproof: 9 Rules for Humans in the Age of Automation by Kevin Roose

"World Economic Forum" Davos, adjacent possible, Airbnb, Albert Einstein, algorithmic bias, algorithmic management, Alvin Toffler, Amazon Web Services, Atul Gawande, augmented reality, automated trading system, basic income, Bayesian statistics, Big Tech, big-box store, Black Lives Matter, business process, call centre, choice architecture, coronavirus, COVID-19, data science, deep learning, deepfake, DeepMind, disinformation, Elon Musk, Erik Brynjolfsson, factory automation, fake news, fault tolerance, Frederick Winslow Taylor, Freestyle chess, future of work, Future Shock, Geoffrey Hinton, George Floyd, gig economy, Google Hangouts, GPT-3, hiring and firing, hustle culture, hype cycle, income inequality, industrial robot, Jeff Bezos, job automation, John Markoff, Kevin Roose, knowledge worker, Kodak vs Instagram, labor-force participation, lockdown, Lyft, mandatory minimum, Marc Andreessen, Mark Zuckerberg, meta-analysis, Narrative Science, new economy, Norbert Wiener, Northpointe / Correctional Offender Management Profiling for Alternative Sanctions, off-the-grid, OpenAI, pattern recognition, planetary scale, plutocrats, Productivity paradox, QAnon, recommendation engine, remote working, risk tolerance, robotic process automation, scientific management, Second Machine Age, self-driving car, Shoshana Zuboff, Silicon Valley, Silicon Valley startup, social distancing, Steve Jobs, Stuart Kauffman, surveillance capitalism, tech worker, The Future of Employment, The Wealth of Nations by Adam Smith, TikTok, Travis Kalanick, Uber and Lyft, uber lyft, universal basic income, warehouse robotics, Watson beat the top human players on Jeopardy!, work culture

Many of us are eminently automatable, especially those of us whose output tends to be more routine and predictable. In 2020, several publications began experimenting with GPT-3, an advanced AI program developed by the nonprofit research lab OpenAI. The program, which takes a prompt and uses machine learning to complete it, was able to produce long, cogent pieces of writing that amazed human editors with their clarity and style. One publication, The Guardian, used GPT-3 to write an entire op-ed about the future of AI and machine learning, and concluded that “overall, it took less time to edit than many human op-eds.” This isn’t to say that machines will replace all white-collar workers, or even most of them.

…

In a 2018 study, twenty top U.S. corporate lawyers Jonathan Marciano, “20 Top Lawyers Were Beaten by Legal AI. Here Are Their Surprising Responses,” Hacker Noon, October 25, 2018. In 2017, Google released AutoML Tom Simonite, “Google’s AI Experts Try to Automate Themselves,” Wired, April 16, 2019. “overall, it took less time” GPT-3, “A Robot Wrote This Entire Article. Are You Scared Yet, Human?,” The Guardian, September 8, 2020. Stanford researchers recently developed Woebot Megan Molteni, “The Chatbot Therapist Will See You Now,” Wired, June 7, 2017. Early research on the effectiveness of eldercare robots Mikaela Law et al., “Developing Assistive Robots for People with Mild Cognitive Impairment and Mild Dementia: A Qualitative Study with Older Adults and Experts in Aged Care,” BMJ Open (2019).

The Internet Is Not What You Think It Is: A History, a Philosophy, a Warning by Justin E. H. Smith

3D printing, Ada Lovelace, Adrian Hon, agricultural Revolution, algorithmic management, artificial general intelligence, Big Tech, Charles Babbage, clean water, coronavirus, COVID-19, cryptocurrency, dark matter, disinformation, Donald Trump, drone strike, Elon Musk, game design, gamification, global pandemic, GPT-3, Internet of things, Isaac Newton, Jacquard loom, Jacques de Vaucanson, Jaron Lanier, jimmy wales, Joseph-Marie Jacquard, Kuiper Belt, Mark Zuckerberg, Marshall McLuhan, meme stock, new economy, Nick Bostrom, Norbert Wiener, packet switching, passive income, Potemkin village, printed gun, QAnon, Ray Kurzweil, Republic of Letters, Silicon Valley, Skype, strong AI, technological determinism, theory of mind, TikTok, Tragedy of the Commons, trolley problem, Turing machine, Turing test, you are the product

As Alex Garland’s remarkable 2014 film, Ex Machina, conveys—updating a timeless plot conceit from Charles Perrault’s folktale Bluebeard—a robot that gains a sense of self, a will, and a consciousness as a result of its complex learning ability, is one that is today best imagined as learning from the totality of data floating around out there, in text messages, chatrooms, search engines, and so on. When we move back from science fiction into reality, this is also how the language-learning neural network known as GPT-3 has recently been able to master human-like language and reasoning with such uncanny perfection. Moreover, those who maintain that human existence might well be a video game–like simulation would likely not have come to think this way if the video games in question were, say, arcade consoles featuring Pac-Man or Space Invaders.

…

See empathy empathy, 25 Enya, 48 Facebook, 1, 43–44, 51, 122, 155, 164 Fage, Robert, 109 filaments, 146; galaxy, 148; mycorhizal, 67 Flaubert, Gustave, 12 Foucault, Michel, 12 freedom, 10–11 Ganeri, Jonardon, 24–28 Garland, Alex, 121 Gass, William, 153 Google, 155 Google Scholar, 41 Gould, Stephen Jay, 71 GPT-3, 121–22 Grindr, 21 Grollier de Servière, Nicolas, 170 Grosseteste, Robert, 101 Gruzinski, Serge, 75 Guattari, Félix, 67 Gutenberg, Johannes, 33 Hacking, Ian, 12, 165–66 Haugeland, John, 96 Hawkins, Screamin’ Jay, 47 Heidegger, Martin, 6, 27 Heine, Heinrich, 75 Herder, J. G., 26 Hippocratics, 60 homunuculus fallacy, 24 Holiday, Billie, 47 Hon, Adrian, 44 HTML, 8 Hughes, Don, 14, 45 Hugo, Victor, 65 Industrial Revolution, 5 Inspector Gadget, 43 Instagram, 36 intelligence, artificial.

Human Frontiers: The Future of Big Ideas in an Age of Small Thinking by Michael Bhaskar

"Margaret Hamilton" Apollo, 3D printing, additive manufacturing, AI winter, Albert Einstein, algorithmic trading, AlphaGo, Anthropocene, artificial general intelligence, augmented reality, autonomous vehicles, backpropagation, barriers to entry, basic income, behavioural economics, Benoit Mandelbrot, Berlin Wall, Big bang: deregulation of the City of London, Big Tech, Bletchley Park, blockchain, Boeing 747, brain emulation, Brexit referendum, call centre, carbon tax, charter city, citizen journalism, Claude Shannon: information theory, Clayton Christensen, clean tech, clean water, cognitive load, Columbian Exchange, coronavirus, cosmic microwave background, COVID-19, creative destruction, CRISPR, crony capitalism, cyber-physical system, dark matter, David Graeber, deep learning, DeepMind, deindustrialization, dematerialisation, Demis Hassabis, demographic dividend, Deng Xiaoping, deplatforming, discovery of penicillin, disruptive innovation, Donald Trump, double entry bookkeeping, Easter island, Edward Jenner, Edward Lorenz: Chaos theory, Elon Musk, en.wikipedia.org, endogenous growth, energy security, energy transition, epigenetics, Eratosthenes, Ernest Rutherford, Eroom's law, fail fast, false flag, Fellow of the Royal Society, flying shuttle, Ford Model T, Francis Fukuyama: the end of history, general purpose technology, germ theory of disease, glass ceiling, global pandemic, Goodhart's law, Google Glasses, Google X / Alphabet X, GPT-3, Haber-Bosch Process, hedonic treadmill, Herman Kahn, Higgs boson, hive mind, hype cycle, Hyperloop, Ignaz Semmelweis: hand washing, Innovator's Dilemma, intangible asset, interchangeable parts, Internet of things, invention of agriculture, invention of the printing press, invention of the steam engine, invention of the telegraph, invisible hand, Isaac Newton, ITER tokamak, James Watt: steam engine, James Webb Space Telescope, Jeff Bezos, jimmy wales, job automation, Johannes Kepler, John von Neumann, Joseph Schumpeter, Kenneth Arrow, Kevin Kelly, Kickstarter, knowledge economy, knowledge worker, Large Hadron Collider, liberation theology, lockdown, lone genius, loss aversion, Louis Pasteur, Mark Zuckerberg, Martin Wolf, megacity, megastructure, Menlo Park, Minecraft, minimum viable product, mittelstand, Modern Monetary Theory, Mont Pelerin Society, Murray Gell-Mann, Mustafa Suleyman, natural language processing, Neal Stephenson, nuclear winter, nudge unit, oil shale / tar sands, open economy, OpenAI, opioid epidemic / opioid crisis, PageRank, patent troll, Peter Thiel, plutocrats, post scarcity, post-truth, precautionary principle, public intellectual, publish or perish, purchasing power parity, quantum entanglement, Ray Kurzweil, remote working, rent-seeking, Republic of Letters, Richard Feynman, Robert Gordon, Robert Solow, secular stagnation, shareholder value, Silicon Valley, Silicon Valley ideology, Simon Kuznets, skunkworks, Slavoj Žižek, sovereign wealth fund, spinning jenny, statistical model, stem cell, Steve Jobs, Stuart Kauffman, synthetic biology, techlash, TED Talk, The Rise and Fall of American Growth, the scientific method, The Wealth of Nations by Adam Smith, Thomas Bayes, Thomas Kuhn: the structure of scientific revolutions, Thomas Malthus, TikTok, total factor productivity, transcontinental railway, Two Sigma, Tyler Cowen, Tyler Cowen: Great Stagnation, universal basic income, uranium enrichment, We wanted flying cars, instead we got 140 characters, When a measure becomes a target, X Prize, Y Combinator

Fusion scientists are optimistic that the application of AI could bring decisive advances in the coming years, and in general the field is now focused on ML approaches to core problems.25 Breakthroughs in natural language processing are coming at pace: the parameters of OpenAI's eye-catching GPT language prediction system grew from hundreds of millions to hundreds of billions in just a few years with some spectacular results, enabling it to write convincing text at length on any subject.26 GPT-3 can take a portion of writing and then continue it with at times shocking plausibility. This is a powerful real-world application already throwing up startling ideas. If you have a chance to play with such generators, nothing so immediately conveys the speed and potential of this new age. As sophistication and computational capacity grow, so does our ability to see new things in the data, beyond the limits of human perception.

…

Some years ago researchers at Aberystwyth and Cambridge Universities developed a robot called Adam, arguably the first automaton scientist capable of formulating hypotheses, running experiments and interpreting the results: studying the metabolism of yeast, and using automated lab equipment, it identified genes that coded for certain enzymes in the yeast.29 Like his biblical equivalent, Adam is only the beginning. AI roams far beyond science and engineering problems. Creativity was always the Promethean province of humans. But for how much longer? Machines convincingly recreate the music of J.S. Bach, the paintings of Rembrandt and the prose style of Tolkien (or BuzzFeed). AI, including GPT-3 mentioned above, produces cogent original texts, whether poetry, journalism or one-liner jokes, perfectly realised images and compelling music. Some of this simply copies humans, but some shows what is possible when the bounds of human creativity are left behind.30 Such work is of real aesthetic interest and value: pieces of music that last and evolve over thousands of years, or images that defy the imagination.

…

., p. 181 106 Rifkin (2014), pp. 82–3 107 Cowen (2018a) 108 Storrs Hall (2018) 7 THE WORLD’S NEW TOOLKIT 1 Goldin and Kutarna (2017), p. 186 2 Senior et al. (2020) 3 Ibid. 4 The founder of CASP prefers to call it an experiment rather than a competition, even if the competitive edge is a large part of what makes it work. 5 AlQuraishi (2018) 6 Reynolds (2020) 7 McAfee and Brynjolfsson (2017), p. 2 8 Callaway (2020) 9 AlQuraishi (2020) 10 Hassabis (2019) 11 Wootton (2015) 12 Ibid. 13 Eisenstein (1979) 14 Wootton (2015), p. 215 15 Although the discovery of sunspots seems to be another example of multiple discovery, also found at the same time by astronomers in Oxford, Ingolstadt and Wittenberg. 16 Galileo quoted in Koestler (1964), p. 336 17 Wootton (2015), p. 236 18 This point is widely made – see for example Mulgan (2018), Agar (2012) or Ridley (2016) for wider discussions. 19 Tools are not then sufficient conditions but they are (often) necessary conditions. 20 https://hbr.org/podcast/2020/10/deepminds-journey-from-games-to-fundamental-science 21 https://home.cern/science/computing/processing-what-record 22 Goldin and Kutarna (2017), p. 241 23 Ringel et al. (2020) 24 Stokes (2020) 25 McMahon (2020) 26 Even that is quickly dated: since releasing GPT-3 Google has a model with trillions of parameters, and the numbers will keep growing. 27 Tshitoyan et al. (2019) 28 Rotman (2019) 29 Malone (2018), p. 240 30 Du Sautoy (2019), Miller (2019) 31 Hafner et al. (2019) 32 Quoted in Parker (2020) 33 Assael, Sommerschield and Prag (2019) 34 Each is a genuine application of AI and not just a series of cool things. 35 See Kelly (2017), pp. 45–6 for a taxonomy of possible minds. 36 Bostrom (2017) 37 Lovelock (2019) 38 Experiments on B-mesons and Muon g-2 respectively.

Exponential: How Accelerating Technology Is Leaving Us Behind and What to Do About It by Azeem Azhar

"Friedman doctrine" OR "shareholder theory", "World Economic Forum" Davos, 23andMe, 3D printing, A Declaration of the Independence of Cyberspace, Ada Lovelace, additive manufacturing, air traffic controllers' union, Airbnb, algorithmic management, algorithmic trading, Amazon Mechanical Turk, autonomous vehicles, basic income, Berlin Wall, Bernie Sanders, Big Tech, Bletchley Park, Blitzscaling, Boeing 737 MAX, book value, Boris Johnson, Bretton Woods, carbon footprint, Chris Urmson, Citizen Lab, Clayton Christensen, cloud computing, collective bargaining, computer age, computer vision, contact tracing, contact tracing app, coronavirus, COVID-19, creative destruction, crowdsourcing, cryptocurrency, cuban missile crisis, Daniel Kahneman / Amos Tversky, data science, David Graeber, David Ricardo: comparative advantage, decarbonisation, deep learning, deglobalization, deindustrialization, dematerialisation, Demis Hassabis, Diane Coyle, digital map, digital rights, disinformation, Dissolution of the Soviet Union, Donald Trump, Double Irish / Dutch Sandwich, drone strike, Elon Musk, emotional labour, energy security, Fairchild Semiconductor, fake news, Fall of the Berlin Wall, Firefox, Frederick Winslow Taylor, fulfillment center, future of work, Garrett Hardin, gender pay gap, general purpose technology, Geoffrey Hinton, gig economy, global macro, global pandemic, global supply chain, global value chain, global village, GPT-3, Hans Moravec, happiness index / gross national happiness, hiring and firing, hockey-stick growth, ImageNet competition, income inequality, independent contractor, industrial robot, intangible asset, Jane Jacobs, Jeff Bezos, job automation, John Maynard Keynes: Economic Possibilities for our Grandchildren, John Maynard Keynes: technological unemployment, John Perry Barlow, Just-in-time delivery, Kickstarter, Kiva Systems, knowledge worker, Kodak vs Instagram, Law of Accelerating Returns, lockdown, low skilled workers, lump of labour, Lyft, manufacturing employment, Marc Benioff, Mark Zuckerberg, megacity, Mitch Kapor, Mustafa Suleyman, Network effects, new economy, NSO Group, Ocado, offshore financial centre, OpenAI, PalmPilot, Panopticon Jeremy Bentham, Peter Thiel, Planet Labs, price anchoring, RAND corporation, ransomware, Ray Kurzweil, remote working, RFC: Request For Comment, Richard Florida, ride hailing / ride sharing, Robert Bork, Ronald Coase, Ronald Reagan, Salesforce, Sam Altman, scientific management, Second Machine Age, self-driving car, Shoshana Zuboff, Silicon Valley, Social Responsibility of Business Is to Increase Its Profits, software as a service, Steve Ballmer, Steve Jobs, Stuxnet, subscription business, synthetic biology, tacit knowledge, TaskRabbit, tech worker, The Death and Life of Great American Cities, The Future of Employment, The Nature of the Firm, Thomas Malthus, TikTok, Tragedy of the Commons, Turing machine, Uber and Lyft, Uber for X, uber lyft, universal basic income, uranium enrichment, vertical integration, warehouse automation, winner-take-all economy, workplace surveillance , Yom Kippur War

Scientists rushed to build artificial intelligence systems, applying deep neural networks and their derivatives to a vast array of problems: from spotting manufacturing defects to translating between languages; from voice recognition to detecting credit card fraud; from discovering new medicines to recommending the next video we should watch. Investors opened their pocketbooks eagerly to back these inventors. In short order, deep learning was everywhere. As a result, neural networks demanded increasing amounts of data and processing power. A 2020 neural net, GPT-3 – used to generate text that could sometimes pass for being written by a human – had 175 billion parameters, about 3,000 times more than AlexNet. But if the new approach to computing was AI, what was powering it? Between 2012 and 2018, the amount of computer power used to train the largest AI models increased about six times faster than the rate of Moore’s Law.

…

Abu Dhabi, UAE, 250 Acemoglu, Daron, 139 Acorn Computers, 16, 21 Ada Lovelace Institute, 8 additive manufacturing, 43–4, 46, 48, 88, 166, 169, 175–9 Adidas, 176 advertising, 94, 112–13, 116, 117, 227–8 AdWords, 227 aeroponics, 171 Afghanistan, 38, 205 Africa, 177–8, 182–3 Aftenposten, 216 Age of Spiritual Machines, The (Kurzweil), 77 agglomeration, 181 Air Jordan sneakers, 102 Airbnb, 102, 188 aircraft, 49–50 Alexandria, Egypt, 180 AlexNet, 33 Algeciras, HMM 61 Alibaba, 48, 102, 108, 111, 122 Alipay, 111 Allen, Robert, 80 Alphabet, 65, 113–14, 131, 163 aluminium, 170 Amazon, 65, 67–8, 94, 104, 108, 112, 122, 135–6 Alexa, 25, 117 automation, 135–6, 137, 139, 154 collective bargaining and, 163 Covid-19 pandemic (2020–21), 135–6 drone sales, 206 Ecobee and, 117 Go stores, 136 Kiva Systems acquisition (2012), 136 management, 154 Mechanical Turk, 142–3, 144, 145 monopoly, 115, 117, 122 Prime, 136, 154 R&D, 67–8, 113 Ami Pro, 99 Amiga, 16 Anarkali, Lahore, 102 anchoring bias, 74 Android, 85, 94, 117, 120 Angola, 186 Ant Brain, 111 Ant Financial, 111–12 antitrust laws, 114, 119–20 Apache HTTP Server, 242 Appelbaum, Binyamin, 63 Apple, 47, 62, 65, 85, 94, 104, 108, 112, 122 App Store, 105, 112, 115 chip production, 113 Covid-19 pandemic (2019–21), 222–3 data collection, 228 iOS, 85 iPhone, 47, 62, 85, 94, 105 media subscription, 112 watches, 112 APT33 hacker group, 198 Aral, Sinan, 238 Aramco, 108, 198 Armenia, 206–7 Arthur, William Brian, 110, 123 artificial intelligence, 4, 8, 31–4, 54, 88, 113, 249 academic brain drain, 118 automation, 125–42 data and, 31–2, 142 data network effect, 106–7 drone technology and, 208, 214 education and, 88 employment and, 126–7 healthcare and, 88, 103 job interviews and, 153 regulation of, 187, 188 arXiv, 59 Asana, 151 Asian Development Bank, 193 Aslam, Yaseen, 148 Assembly Bill 5 (California, 2019), 148 asymmetric conflict, 206 AT&T, 76, 100 Atari, 16 attack surfaces, 192–3, 196, 209, 210 Aurora, 141 Australia, 102, 197 automation, 125–42 autonomous weapons, 208, 214 Azerbaijan, 173, 206–7 Ballmer, Steve, 85 Bangladesh, 175 banking, 122, 237 Barcelona, Catalonia, 188 Barlow, John Perry, 184 Barrons, Richard, 195, 211 Bartlett, Albert, 73 batteries, 40, 51, 53–4, 250, 251 Battle of the Overpass (1937), 162 Bayraktar TB2 drone, 206 Bee Gees, 72 Bekar, Clifford, 45 Bell Labs, 18 Bell Telephone Company, 100 Benioff, Marc, 108–9 Bentham, Jeremy, 152 Berlin Wall, fall of (1989), 4 Bermuda, 119 Berners-Lee, Timothy, 55, 100, 160, 239 Bessen, James, 46 Bezos, Jeffrey, 135–6 BGI, 41 Biden, Joseph, 225 Bing, 107 biological weapons, 207, 213 biology, 10, 39, 40–42, 44, 46 genome sequencing, 40–41, 90, 229, 234, 245–7, 250, 252 synthetic biology, 42, 46, 69, 174, 245, 250 biopolymers, 42 bits, 18 Black Death (1346–53), 12 BlackBerry, 120 Blair, Tony, 81 Bletchley Park, Buckinghamshire, 22 blitzscaling, 110 Blockbuster, 138 BMW, 177 Boeing, 51, 236 Bol.com, 103 Bollywood, 181 Boole, George, 18 Bork, Robert, 114–15, 117, 119 Bosworth, Andrew, 233 Boyer, Pascal, 75 Boyle, James, 234 BP, 92, 158 brain, 77 Braudel, Fernand, 75 Brave, 242 Brazil, 202 Bremmer, Ian, 187 Bretton Woods Conference (1944), 87 Brexit (2016–20), 6, 168 British Broadcasting Corporation (BBC), 87, 129, 191 Brookings Institution, 130 BT, 123 Bulgaria, 145 Bundy, Willard Legrand, 149 Busan, South Korea, 56 business, 82, 92–124 diminishing returns to scale, 93, 108 economic dynamism and, 117 economies of scale, 50, 92 growth, 110–13 increasing returns to scale, 108–10 intangible economy, 104–7, 118, 156, 175, 180 linear value chains, 101 market share, 93–6, 111 monopolies, 10, 71, 94, 95, 114–24 network effect, 96–101 platform model, 101–3, 219 re-localisation, 11, 166–79, 187, 252, 255 state-sized companies, 11, 67 superstar companies, 10, 94–6 supply chains, 61–2, 166–7, 169, 175, 187, 252, 255 taxation of, 96, 118–19 Butler, Nick, 179 ByteDance, 28 C40 initiative, 189 Cambridge University, 127, 188 cancer, 57–8, 127 Capitol building storming (2021), 225 car industry, 93 carbon emissions, 35, 90, 251 Carlaw, Kenneth, 45 Carnegie, Andrew, 112 Carnegie Mellon University, 131 Catholic Church, 83, 88 censorship, 216–17, 224–6, 236 Central Intelligence Agency (CIA), 194 Cerebras, 34 cervical smears, 57–8 chemical weapons, 207, 213 Chen, Brian, 228 chewing gum, 78 Chicago Pile-1 reactor, 64 Chile, 170 China automation in, 127, 137 brainwave reading in, 152 Covid-19 pandemic (2019–21), 245 drone technology in, 207 Great Firewall, 186, 201 Greater Bay Area, 182 horizontal expansion in, 111–12 manufacturing in, 176 misinformation campaigns, 203 raw materials, demand for, 178 Singles’ Day, 48 social credit systems, 230 superstar companies in, 95 US, relations with, 166 chips, 19–22, 28–9, 48–9, 52, 113, 251 Christchurch massacre (2019), 236 Christensen, Clayton, 24 CIPD, 153 cities, 11, 75, 169, 179–84, 188, 255 Clegg, Nick, 225–6, 235 climate change, 90, 169, 187, 189, 251, 252 cloud computing, 85, 112 Cloudflare, 200 cluster bombs, 213 CNN, 185, 190 coal, 40, 65, 172 Coase, Ronald, 92 Coca-Cola, 93 code is law, 220–22, 235 cold fusion, 113–14 Cold War (1947–91), 194, 212, 213 collective bargaining, 147, 149, 154, 156, 162–5 Colombia, 145 colonialism, 167 Columbus, Christopher, 4 combination, 53–7 Comical Ali, 201 commons, 234–5, 241–3, 256 companies, see business comparative advantage, 170 complex systems, 2 compounding, 22–3, 28 CompuServe, 100 computing, 4, 10, 15–36, 44, 46, 249 artificial intelligence, 4, 8, 31–4, 54, 88 cloud computing, 85, 112 internet, 47–8, 55, 65, 84 Law of Accelerating Returns, 30–31, 33, 35 machining, 43 Moore’s Law, see Moore’s Law quantum computing, 35 transistors, 18–22, 28–9, 48–9, 52 conflict, 87, 189, 190–215 attack surfaces, 192–3, 196, 209, 210 cyberattacks, 11, 114, 140, 181, 187, 190–200, 209–14, 256 de-escalation, 212–13 drone technology, 11, 192, 204–9, 214, 256 institutional change and, 87 misinformation, 11, 191, 192, 200–204, 209, 212, 217, 225 new wars, 194 non-proliferation, 213–14 re-localisation and, 189, 193, 194, 209 consent of the networked, 223 Costco, 67 Coursera, 58 Covid-19 pandemic (2019–21), 12–13, 59, 78–9, 131, 245–9 automation and, 127, 135, 136 cities and, 183 contact-tracing apps, 222–3 gig economy and, 146 lockdowns, 12, 152, 176, 183, 246 manufacturing and, 176 misinformation and, 202–4, 247–8 preprint servers and, 60 recession (2020–21), 178 remote working and, 146, 151, 153 supply chains and, 169, 246 vaccines, 12, 202, 211, 245–7 workplace cultures and, 151, 152 cranks, 54 credit ratings, 162, 229 critical thinking skills, 212 Croatia, 145 Crocker, David, 55 crowdsourcing, 143–4 Cuba, 203 Cuban missile crisis (1962), 99, 212 cultural lag, 85 cyberattacks, 11, 114, 140, 181, 187, 190–200, 209–14, 256 CyberPeace Institute, 214 Daniel, Simon, 173–4 Dar es Salaam, Tanzania, 183 Darktrace, 197 data, 8, 11, 71, 217–19, 226–31, 235, 237–42, 256 AI and, 8, 32, 33, 58, 106 compensation for, 239 commons, 242 cyberattacks and, 196 doppelgängers, 219, 226, 228, 239 interoperability and, 237–9 network effects, 106–7, 111 protection laws, 186, 226 rights, 240 Daugherty, Paul, 141 DDT (dichlorodiphenyltrichloroe thane), 253 death benefits, 151 Dediu, Horace, 24, 30 deep learning, 32–4, 54, 58, 127 deforestation, 251 dehumanisation, 71, 154, 158 deindustrialisation, 168 Deliveroo, 154, 163 Delphi, 100 dematerialised techniques, 166, 175 Denmark, 58, 160, 199–200, 257 Deutsche Bank, 130 Diamandis, Peter, 5 Dickens, Charles, 80 digital cameras, 83–4 Digital Geneva Convention, 211 Digital Markets Act (EU, 2020), 122 digital minilateralism, 188 Digital Nations group, 188 Digital Services Act (EU, 2020), 123 diminishing returns, 93, 108 disinformation, see misinformation DoorDash, 147, 148, 248 dot-com bubble (1995–2000), 8, 108, 150 Double Irish tax loophole, 119 DoubleClick, 117 drone technology, 11, 192, 204–9, 214, 256 Dubai, UAE, 43 Duke University, 234 dystopia, 208, 230, 253 Eagan, Nicole, 197 eBay, 98, 121 Ecobee, 120 economies of scale, 50, 92 Economist, The, 8, 65, 119, 183, 239 economists, 63 Edelman, 3 education artificial intelligence and, 88 media literacy, 211–12 Egypt, 145, 186 Elance, 144 electric cars, 51, 69, 75, 173–4, 177, 250 electricity, 26, 45, 46, 54, 157, 249–50 see also energy Electronic Frontier Foundation, 184 email, 6, 55 embodied institutions, 82 employment, 10, 71, 125–65 automation, 125–42 collective bargaining, 147, 149, 154, 156, 162–5 dehumanisation and, 71, 154, 158 flexicurity, 160–61, 257 gig economy, 10, 71, 142–9, 153, 162, 164, 239, 252, 255 income inequality, 155–8, 161, 168 lump of labour fallacy, 139 management, 149–54, 158–9 protections, 85–6, 147–9 reskilling, 159–60 universal basic income (UBI), 160, 189 Enclosure, 234–5, 241 energy, 11, 37–8, 39–40, 44, 46, 172–4, 250 cold fusion, 113–14 fossil fuels, 40, 159, 172, 250 gravitational potential, 53 solar power, 37–8, 53, 65, 77, 82, 90, 171, 172, 173, 249, 250, 251 storage, 40, 53, 114, 173–4, 250, 251 wind power, 39–40, 52 Energy Vault, 53–4, 173 Engels, Friedrich, 81 Engels’ pause, 80, 81 environmental movement, 73 Epic Games, 116 estate agents, 100 Estonia, 188, 190–91, 200, 211 Etzion Airbase, Sinai Peninsula, 195 European Commission, 116, 122, 123 European Space Agency, 56 European Union, 6, 82, 147, 186, 226 Excel, 99 exogeny, 2 exponential gap, 9, 10, 67–91, 70, 89, 253 cyber security and, 193 institutions and, 9, 10, 79–88, 90 mathematical understanding and, 71–5 predictions and, 75–9 price declines and, 68–9 superstar companies and, 10, 94–124 exponential growth bias, 73 Exponential View, 8–9 externalities, 97 extremism, 232–4 ExxonMobil, 65, 92 Facebook, 27, 28, 65, 94, 104, 108, 122, 216–17, 218, 219, 221–2, 223 advertising business, 94, 228 censorship on, 216–17, 224–6, 236 collective bargaining and, 164 data collection on, 228, 239–40 extremism and, 233–4 Instagram acquisition (2012), 117, 120 integrity teams, 234 interoperability, 237–8 Kenosha unrest shooting (2020), 224 misinformation on, 201, 225 network effect and, 98, 223 Oculus acquisition (2014), 117 pay at, 156–7 Phan photo controversy (2016), 216–17, 224, 225 platform model, 101 polarisation and, 233 relationship status on, 221–2 Rohingya ethnic cleansing (2018), 224, 225 US presidential election (2016), 217 WhatsApp acquisition (2014), 117 facial recognition, 152, 208 Factory Act (UK, 1833), 81 Fairchild Semiconductor, 19, 21 fake news, 201–4 family dinners, 86 farming, 170–72, 251 Farrar, James, 148 fax machines, 97 Federal Aviation Administration (US), 236 feedback loops, 3, 13 fertilizers, 35, 90 5G, 203 Financial Conduct Authority, 122 Financial Times, 183 Finland, 160, 211–12 Fitbit, 158 Fiverr, 144 flashing of headlights, 83 flexicurity, 160, 257 flints, 42 flywheels, 54 Ford, 54, 92, 162 Ford, Gerald, 114 Ford, Henry, 54, 162 Ford, Martin, 125 Fortnite, 116 fossil fuels, 40, 159, 172 France, 100, 138, 139, 147, 163 free-market economics, 63–4 freelance work, 10, 71, 142–9 Frey, Carl, 129, 134, 141 Friedman, Milton, 63–4, 241 Friedman, Thomas, 167 FriendFeed, 238 Friendster, 26 Fudan University, 245 fund management, 132 Galilei, Galileo, 83 gaming, 86 Gates, Bill, 17, 25, 84 gender, 6 General Agreement on Tariffs and Trade, 87 General Data Protection Regulation (GDPR), 226 General Electric, 52 General Motors, 92, 125, 130 general purpose technologies, 10, 45–8 generative adversarial networks (GANs), 58 Geneva Conventions, 193, 199, 209 Genghis Khan, 44 GEnie, 100 genome sequencing, 40–41, 90, 229, 234, 245–7, 250, 252 Germany, 75, 134, 147 Giddens, Anthony, 82 gig economy, 10, 71, 142–9, 153, 162, 164, 239, 252, 255 Gilbreth, Lillian, 150 Ginsparg, Paul, 59 GitHub, 58, 60 GlaxoSmithKline, 229–30 global financial crisis (2007–9), 168 Global Hawk drones, 206 global positioning systems (GPS), 197 globalisation, 11, 62, 64, 156, 166, 167–71, 177, 179, 187, 193 internet and, 185 conflict and, 189, 193, 194 Glocer, Thomas, 56 Go (game), 132 GOAT, 102 Gojek, 103 Golden Triangle, 170 Goldman Sachs, 151 Goodfellow, Ian, 58 Google, 5, 35, 36, 94, 98, 104, 108, 115, 122 advertising business, 94, 112–13, 116, 117, 227 Android, 85, 94, 117, 120 chip production, 113 Covid-19 pandemic (2019–21), 222–3 data network effect, 106–7 death benefits, 151 Double Irish tax loophole, 119 Maps, 113 quantum computing, 35 R&D, 114, 118 vertical integration, 112–13, 116 X, 114 YouTube acquisition (2006), 112, 117 Gopher, 59, 100 GPT-3, 33 Graeber, David, 133–4 Grand Bazaar, Istanbul, 102 Graphcore, 34, 35 graphics chips, 34 Grateful Dead, The, 184 gravitational potential energy, 53 gravity bombs, 195 Greater Bay Area, China, 182 Greenberg, Andy, 199 Gross, Bill, 53 Grove, Andrew, 17 GRU (Glavnoje Razvedyvatel’noje Upravlenije), 199 Guangzhou, Guangdong, 182 Guardian, 8, 125, 154, 226, 227 Guiyang, Guizhou, 166 H1N1 virus, 75 Habermas, Jürgen, 218 Hard Times (Dickens), 80 Hardin, Garrett, 241 Harop drones, 207–8 Harpy drones, 207–8 Harvard University, 150, 218, 220, 221, 253 healthcare artificial intelligence and, 57–8, 88, 103 data and, 230, 239, 250–51 wearable devices and, 158, 251 Helsinki, Finland, 160 Herlev Hospital, Denmark, 58 Hinton, Geoffrey, 32, 126–7 HIPA Act (US, 1996), 230 Hitachi, 152 Hobbes, Thomas, 210 Hoffman, Josh, 174 Hoffman, Reid, 110, 111 Holmes, Edward, 245 homophily, 231–4 Hong Kong, 182 horizontal expansion, 111–12, 218 Houston Islam protests (2016), 203 Houthis, 206 Howe, Jeff, 143 Hsinchu, Taiwan, 181 Hughes, Chris, 217 Hull, Charles, 43 Human + Machine (Daugherty), 141 human brain, 77 human genome, 40–41, 90, 229, 234, 250 human resources, 150 Hussein, Saddam, 195 Hyaline, 174 hydroponics, 171 hyperinflation, 75 IBM, 17, 21, 47, 98 IDC, 219 Ideal-X, 61 Ikea, 144 Illumina, 41 Ilves, Toomas Hendrik, 190 ImageNet, 32 immigration, 139, 168, 183–4 Impossible Foods, 69 Improv, 99 income inequality, 155–8, 161, 168 India, 103, 145, 181, 186, 224, 253, 254 Indonesia, 103 Industrial Revolution (1760–1840), 79–81, 157, 235 informational networks, 59–60 ING, 178 innovation, 14, 117 Innovator’s Dilemma, The (Christensen), 24 Instagram, 84, 117, 120, 121, 237 institutions, 9, 10, 79–88, 90–91 path dependence, 86–7 punctuated equilibrium, 87–8 intangible economy, 104–7, 118, 156, 175, 180 integrated circuits, 19 Intel, 16–17, 19, 163 intellectual property law, 82 Intermediate-Range Nuclear Forces Treaty (1987), 237 International Alliance of App-Based Transport Workers, 164 International Court of Justice, 224 International Criminal Court, 208 International Energy Agency, 77, 82 International Labour Organization, 131 International Monetary Fund (IMF), 87, 167, 187 international organisations, 82 International Organization for Standardization, 55, 61 International Rescue Committee, 184 International Telecommunication Union, 55 internet, 7, 47–8, 55, 65, 72, 75, 84–5, 88, 115, 184–6 code is law, 220–22, 235 data and, 11, 32, 71 informational networks, 59–60 localisation, 185–6 lockdowns and, 12 network effect, 100–101 online shopping, 48, 61, 62, 75, 94, 102, 135 platform model and, 102 public sphere and, 223 standardisation, 55 Wi-Fi, 151 interoperability, 55, 120–22, 237–9, 241, 243, 256–7 iPhone, 47, 62, 85, 94, 115, 175 Iran, 186, 196, 198, 203, 206 Iraq, 195–6, 201, 209 Ireland, 57–8, 119 Islamic State, 194, 233 Israel, 37, 188, 195–6, 198, 206, 207–8 Istanbul, Turkey, 102 Jacobs, Jane, 182 Japan, 37, 152, 171, 174 Jasanoff, Sheila, 253 JD.com, 137 Jena, Rajesh, 127 Jio, 103 job interviews, 153, 156 John Paul II, Pope, 83 Johnson, Boris, 79 Jumia, 103 just in time supply chains, 61–2 Kahneman, Daniel, 74 KakaoTalk, 27 Kaldor, Mary, 194 Kapor, Mitchell, 99 Karunaratne, Sid, 140–41, 151 Kenosha unrest shooting (2020), 224 Keynes, John Maynard, 126, 158 Khan, Lina, 119 Khartoum, Sudan, 183 Kim Jong-un, 198 King’s College London, 179 Kiva Systems, 136 Kobo360, 145 Kodak, 83–4, 88 Kranzberg, Melvin, 254 Krizhevsky, Alex, 32–3, 34 Kubursi, Atif, 178 Kurdistan Workers’ Party, 206 Kurzweil, Ray, 29–31, 33, 35, 77 Lagos, Nigeria, 182 Lahore, Pakistan, 102 landmines, 213 Law of Accelerating Returns, 30–31, 33, 35 Laws of Motion, 20 learning by doing, 48, 53 Leggatt, George, 148 Lemonade, 56 Lessig, Larry, 220–21 Leviathan (Hobbes), 210 Li Fei-Fei, 32 life expectancy, 25, 26 light bulbs, 44, 157 Lime, 27 Limits to Growth, The (Meadows et al.), 73 linear value chains, 101 LinkedIn, 26, 110, 121, 237, 238 Linkos Group, 197 Linux OS, 242 Lipsey, Richard, 45 lithium-ion batteries, 40, 51 lithium, 170 localism, 11, 166–90, 252, 255 log files, 227 logarithmic scales, 20 logic gates, 18 logistic curve, 25, 30, 51, 52, 69–70 London, England, 180, 181, 183 London Underground, 133–4 looms, 157 Lordstown Strike (1972), 125 Lotus Development Corporation, 99 Luddites, 125, 253 Lufa Farms, 171–2 Luminate, 240 lump of labour fallacy, 139 Lusaka, Zambia, 15 Lyft, 146, 148 machine learning, 31–4, 54, 58, 88, 127, 129, 143 MacKinnon, Rebecca, 223 Maersk, 197, 199, 211 malaria, 253 Malaysia Airlines Flight 17 shootdown (2014), 199 Malta, 114 Malthus, Thomas, 72–3 malware, 197 Man with the Golden Gun, The (1974 film), 37 manufacturing, 10, 39, 42–4, 46, 166–7, 175–9 additive, 43–4, 46, 48, 88, 166, 169, 175–9 automation and, 130 re-localisation, 175–9 subtractive, 42–3 market saturation, 25–8, 51, 52 market share, 93–6, 111 Marshall, Alfred, 97 Massachusetts Institute of Technology, 18, 147, 202, 238 Mastercard, 98 May, Theresa, 183 Mayors for a Guaranteed Income, 189 McCarthy, John, 31 McKinsey, 76, 94 McMaster University, 178 measles, 246 Mechanical Turk, 142–3, 144, 145 media literacy, 211–12 meningitis, 246 Mexico, 202 microorganisms, 42, 46, 69 Microsoft, 16–17, 65, 84–5, 88, 98–9, 100, 105, 108, 122, 221 Bing, 107 cloud computing, 85 data collection, 228 Excel, 99 internet and, 84–5, 100 network effect and, 99 Office software, 98–9, 110, 152 Windows, 85, 98–9 Workplace Productivity scores, 152 Mill, John Stuart, 193 miniaturisation, 34–5 minimum wage, 147, 161 misinformation, 11, 191, 192, 200–204, 209, 212, 217, 225, 247–8 mobile phones, 76, 121 see also smartphones; telecom companies Moderna, 245, 247 Moixa, 174 Mondelez, 197, 211 Mongol Empire (1206–1368), 44 monopolies, 10, 71, 94, 95, 114–24, 218, 255 Monopoly (board game), 82 Montreal, Quebec, 171 mood detection systems, 152 Moore, Gordon, 19, 48 Moore’s Law, 19–22, 26, 28–9, 31, 34, 63, 64, 74 artificial intelligence and, 32, 33–4 Kodak and, 83 price and, 41–2, 51, 68–9 as social fact, 29, 49 superstar companies and, 95 time, relationship with, 48–9 Moravec, Hans, 131 Moravec’s paradox, 131–2 Motorola, 76 Mount Mercy College, Cork, 57 Mozilla Firefox, 242 Mumbai, India, 181 mumps, 246 muskets, 54–5 MySpace, 26–7 Nadella, Satya, 85 Nagorno-Karabakh War (2020), 206–7 napalm, 216 NASA (National Aeronautics and Space Administration), 56 Natanz nuclear site, Iran, 196 National Health Service (NHS), 87 nationalism, 168, 186 NATO (North Atlantic Treaty Organization), 191, 213 Netflix, 104, 107, 109, 136, 137, 138, 139, 151, 248 Netherlands, 103 Netscape Communicator, 6 networks, 58–62 network effects, 96–101, 106, 110, 121, 223 neural networks, 32–4 neutral, technology as, 5, 220–21, 254 new wars, 194 New York City, New York, 180, 183 New York Times, 3, 125, 190, 228 New Zealand, 188, 236 Newton, Isaac, 20 Nigeria, 103, 145, 182, 254 Niinistö, Sauli, 212 Nike, 102 nitrogen fertilizers, 35 Nixon, Richard, 25, 114 Nobel Prize, 64, 74, 241 Nokia, 120 non-state actors, 194, 213 North Korea, 198 North Macedonia, 200–201 Norway, 173, 216 NotPetya malware, 197, 199–200, 211, 213 Novell, 98 Noyce, Robert, 19 NSO Group, 214 nuclear weapons, 193, 195–6, 212, 237 Nuremberg Trials (1945–6), 208 O’Reilly, Tim, 107 O’Sullivan, Laura, 57–8, 60 Obama, Barack, 205, 214, 225 Ocado, 137 Ocasio-Cortez, Alexandria, 239 Oculus, 117 oDesk, 144 Ofcom, 8 Ofoto, 84 Ogburn, William, 85 oil industry, 172, 250 Houthi drone attacks (2019), 206 OAPEC crisis (1973–4), 37, 258 Shamoon attack (2012), 198 Standard Oil breakup (1911), 93–4 Olduvai, Tanzania, 42 online shopping, 48, 61, 62, 75, 94, 102, 135 open-source software, 242 Openreach, 123 Operation Opera (1981), 195–6, 209 opium, 38 Orange, 121 Organisation for Economic Co-operation and Development (OECD), 119, 167 Osborne Computer Corporation, 16 Osborne, Michael, 129 Osirak nuclear reactor, Iraq, 195–6, 209 Ostrom, Elinor, 241 Oxford University, 129, 134, 203, 226 pace of change, 3 pagers, 87 Pakistan, 145, 205 palladium, 170 PalmPilot, 173 panopticon, 152 Paris, France, 181, 183 path dependence, 86 PayPal, 98, 110 PC clones, 17 PeerIndex, 8, 201, 237 Pegasus, 214 PeoplePerHour, 144 PepsiCo, 93 Perez, Carlota, 46–7 pernicious polarization, 232 perpetual motion, 95, 106, 107, 182 Petersen, Michael Bang, 75 Phan Thi Kim Phuc, 216–17, 224, 225 pharmaceutical industry, 6, 93, 250 phase transitions, 4 Philippines, 186, 203 Phillips Exeter Academy, 150 phishing scams, 211 Phoenix, Arizona, 134 photolithography, 19 Pigou, Arthur Cecil, 97 Piketty, Thomas, 160 Ping An Good Doctor, 103, 250 Pix Moving, 166, 169, 175 PKK (Partîya Karkerên Kurdistanê), 206 Planet Labs, 69 platforms, 101–3, 219 PlayStation, 86 plough, 157 Polanyi, Michael, 133 polarisation, 231–4 polio, 246 population, 72–3 Portify, 162 Postel, Jon, 55 Postings, Robert, 233 Predator drones, 205, 206 preprints, 59–60 price gouging, 93 price of technology, 22, 68–9 computing, 68–9, 191, 249 cyber-weapons, 191–2 drones, 192 genome sequencing, 41–2, 252 renewable energy, 39–40, 250 printing press, 45 public sphere, 218, 221, 223 Pulitzer Prize, 216 punctuated equilibrium, 87–8 al-Qaeda, 205, 210–11 Qatar, 198 quantum computing, 35 quantum physics, 29 quarantines, 12, 152, 176, 183, 246 R&D (research and development), 67–8, 113, 118 racial bias, 231 racism, 225, 231, 234 radicalisation pathways, 233 radiologists, 126 Raford, Noah, 43 Raz, Ze’ev, 195, 209 RB, 197 re-localisation, 11, 166–90, 253, 255 conflict and, 189, 193, 194, 209 Reagan, Ronald, 64, 163 religion, 6, 82, 83 resilience, 257 reskilling, 159–60 responsibility gap, 209 Restrepo, Pascual, 139 Reuters, 8, 56, 132 revolutions, 87 Ricardo, David, 169–70, 177 rights, 240–41 Rise of the Robots, The (Ford), 125 Rittenhouse, Kyle, 224 Roche, 67 Rockefeller, John, 93 Rohingyas, 224 Rome, ancient, 180 Rose, Carol, 243 Rotterdam, Netherlands, 56 Rule of Law, 82 running shoes, 102, 175–6 Russell, Stuart, 31, 118 Russian Federation, 122 disinformation campaigns, 203 Estonia cyberattacks (2007), 190–91, 200 Finland, relations with, 212 Nagorno-Karabakh War (2020), 206 nuclear weapons, 237 Ukraine cyberattacks (2017), 197, 199–200 US election interference (2016), 217 Yandex, 122 S-curve, 25, 30, 51, 52, 69–70 al-Sahhaf, Muhammad Saeed, 201 Salesforce, 108–9 Saliba, Samer, 184 salt, 114 Samsung, 93, 228 San Francisco, California, 181 Sandel, Michael, 218 Sanders, Bernard, 163 Sandworm, 197, 199–200, 211 Santander, 95 Sasson, Steve, 83 satellites, 56–7, 69 Saturday Night Fever (1977 soundtrack), 72 Saudi Arabia, 108, 178, 198, 203, 206 Schmidt, Eric, 5 Schwarz Gruppe, 67 Second Machine Age, The (Brynjolfsson and McAfee), 129 self-driving vehicles, 78, 134–5, 141 semiconductors, 18–22, 28–9, 48–9, 52, 113, 251 September 11 attacks (2001), 205, 210–11 Shamoon virus, 198 Shanghai, China, 56 Shannon, Claude, 18 Sharp, 16 Shenzhen, Guangdong, 182 shipping containers, 61–2, 63 shopping, 48, 61, 62, 75, 94, 102, 135 Siemens, 196 silicon chips, see chips Silicon Valley, 5, 7, 15, 24, 65, 110, 129, 223 Sinai Peninsula, 195 Sinclair ZX81, 15, 17, 21, 36 Singapore, 56 Singles’ Day, 48 Singularity University, 5 SixDegrees, 26 Skydio R1 drone, 208 smartphones, 22, 26, 46, 47–8, 65, 86, 88, 105, 111, 222 Smith, Adam, 169–70 sneakers, 102, 175–6 Snow, Charles Percy, 7 social credit systems, 230 social media, 26–8 censorship on, 216–17, 224–6, 236 collective bargaining and, 164 data collection on, 228 interoperability, 121, 237–8 market saturation, 25–8 misinformation on, 192, 201–4, 217, 247–8 network effect, 98, 223 polarisation and, 231–4 software as a service, 109 solar power, 37–8, 53, 65, 77, 82, 90, 171, 172, 173, 249, 250, 251 SolarWinds, 200 Solberg, Erna, 216 South Africa, 170 South Korea, 188, 198, 202 Southey, Robert, 80 sovereignty, 185, 199, 214 Soviet Union (1922–91), 185, 190, 194, 212 Spain, 170, 188 Spanish flu pandemic (1918–20), 75 Speedfactory, Ansbach, 176 Spire, 69 Spotify, 69 Sputnik 1 orbit (1957), 64, 83 stagflation, 63 Standard and Poor, 104 Standard Oil, 93–4 standardisation, 54–7, 61, 62 Stanford University, 32, 58 Star Wars franchise, 99 state-sized companies, 11, 67 see also superstar companies states, 82 stirrups, 44 Stockholm International Peace Research Institute, 208 Stockton, California, 160 strategic snowflakes, 211 stress tests, 237 Stuxnet, 196, 214 Sudan, 183 superstar companies, 10, 11, 67, 94–124, 218–26, 252, 255 blitzscaling, 110 collective bargaining and, 163 horizontal expansion, 111–12, 218 increasing returns to scale, 108–10 innovation and, 117–18 intangible economy, 104–7, 118, 156 interoperability and, 120–22, 237–9 monopolies, 114–24, 218 network effect, 96–101, 121 platform model, 101–3, 219 taxation of, 118–19 vertical expansion, 112–13 workplace cultures, 151 supply chains, 61–2, 166–7, 169, 175, 187, 252 surveillance, 152–3, 158 Surviving AI (Chace), 129 Sutskever, Ilya, 32 synthetic biology, 42, 46, 69, 174, 245, 250 Syria, 186 Taiwan, 181, 212 Talkspace, 144 Tallinn, Estonia, 190 Tang, Audrey, 212 Tanzania, 42, 183 TaskRabbit, 144 Tasmania, Australia, 197 taxation, 10, 63, 96, 118–19 gig economy and, 146 superstar companies and, 118–19 Taylor, Frederick Winslow, 150, 152, 153, 154 Tel Aviv, Israel, 181 telecom companies, 122–3 Tencent, 65, 104, 108, 122 territorial sovereignty, 185, 199, 214 Tesco, 67, 93 Tesla, 69, 78, 113 Thailand, 176, 203 Thatcher, Margaret, 64, 163 Thelen, Kathleen, 87 Thiel, Peter, 110–11 3D printing, see additive manufacturing TikTok, 28, 69, 159–60, 219 Tisné, Martin, 240 Tomahawk missiles, 207 Toyota, 95 trade networks, 61–2, 166–7, 169, 175 trade unions, see collective bargaining Trading Places (1983 film), 132 Tragedy of the Commons, The (Hardin), 241 transistors, 18–22, 28–9, 48–9, 52, 113, 251 transparency, 236 Treaty of Westphalia (1648), 199 TRS-80, 16 Trump, Donald, 79, 119, 166, 201, 225, 237 Tufekci, Zeynep, 233 Turing, Alan, 18, 22 Turkey, 102, 176, 186, 198, 202, 206, 231 Tversky, Amos, 74 23andMe, 229–30 Twilio, 151 Twitch, 225 Twitter, 65, 201, 202, 219, 223, 225, 237 two cultures, 7, 8 Uber, 69, 94, 102, 103, 106, 142, 144, 145 Assembly Bill 5 (California, 2019), 148 engineering jobs, 156 London ban (2019), 183, 188 London protest (2016), 153 pay at, 147, 156 satisfaction levels at, 146 Uber BV v Aslam (2021), 148 UiPath, 130 Ukraine, 197, 199 Unilever, 153 Union of Concerned Scientists, 56 unions, see collective bargaining United Arab Emirates, 43, 198, 250 United Autoworkers Union, 162 United Kingdom BBC, 87 Biobank, 242 Brexit (2016–20), 6, 168 collective bargaining in, 163 Covid-19 epidemic (2020–21), 79, 203 DDT in, 253 digital minilateralism, 188 drone technology in, 207 flashing of headlights in, 83 Golden Triangle, 170 Google and, 116 Industrial Revolution (1760–1840), 79–81 Luddite rebellion (1811–16), 125, 253 misinformation in, 203, 204 National Cyber Force, 200 NHS, 87 self-employment in, 148 telecom companies in, 123 Thatcher government (1979–90), 64, 163 United Nations, 87, 88, 188 United States antitrust law in, 114 automation in, 127 Battle of the Overpass (1937), 162 Capitol building storming (2021), 225 China, relations with, 166 Cold War (1947–91), 194, 212, 213 collective bargaining in, 163 Covid-19 epidemic (2020–21), 79, 202–4 Cyber Command, 200, 210 DDT in, 253 drone technology in, 205, 214 economists in, 63 HIPA Act (1996), 230 Kenosha unrest shooting (2020), 224 Lordstown Strike (1972), 125 manufacturing in, 130 misinformation in, 202–4 mobile phones in, 76 nuclear weapons, 237 Obama administration (2009–17), 205, 214 polarisation in, 232 presidential election (2016), 199, 201, 217 presidential election (2020), 202–3 Reagan administration (1981–9), 64, 163 self-employment in, 148 September 11 attacks (2001), 205, 210–11 shipping containers in, 61 shopping in, 48 solar energy research, 37 Standard Oil breakup (1911), 93–4 taxation in, 63, 119 Trump administration (2017–21), 79, 119, 166, 168, 201, 225, 237 Vietnam War (1955–75), 216 War on Terror (2001–), 205 universal basic income (UBI), 160, 189 universal service obligation, 122 University of Cambridge, 127, 188 University of Chicago, 63 University of Colorado, 73 University of Delaware, 55 University of Oxford, 129, 134, 203, 226 University of Southern California, 55 unwritten rules, 82 Uppsala Conflict Data Program, 194 UpWork, 145–6 USB (Universal Serial Bus), 51 Ut, Nick, 216 utility providers, 122–3 vaccines, 12, 202, 211, 245–7 Vail, Theodore, 100 value-free, technology as, 5, 220–21, 254 Veles, North Macedonia, 200–201 Véliz, Carissa, 226 Venezuela, 75 venture capitalists, 117 vertical expansion, 112–13, 116 vertical farms, 171–2, 251 video games, 86 Vietnam, 61, 175, 216 Virological, 245 Visa, 98 VisiCalc, 99 Vodafone, 121 Vogels, Werner, 68 Wag!

What We Owe the Future: A Million-Year View by William MacAskill

Ada Lovelace, agricultural Revolution, Albert Einstein, Alignment Problem, AlphaGo, artificial general intelligence, Bartolomé de las Casas, Bletchley Park, British Empire, Brownian motion, carbon footprint, carbon tax, charter city, clean tech, coronavirus, COVID-19, cuban missile crisis, decarbonisation, deep learning, DeepMind, Deng Xiaoping, different worldview, effective altruism, endogenous growth, European colonialism, experimental subject, feminist movement, framing effect, friendly AI, global pandemic, GPT-3, hedonic treadmill, Higgs boson, income inequality, income per capita, Indoor air pollution, Intergovernmental Panel on Climate Change (IPCC), Isaac Newton, Islamic Golden Age, iterative process, Jeff Bezos, job satisfaction, lab leak, Lao Tzu, Large Hadron Collider, life extension, lockdown, long peace, low skilled workers, machine translation, Mars Rover, negative emissions, Nick Bostrom, nuclear winter, OpenAI, Peter Singer: altruism, Peter Thiel, QWERTY keyboard, Robert Gordon, Rutger Bregman, Sam Altman, seminal paper, Shenzhen special economic zone , Shenzhen was a fishing village, Silicon Valley, special economic zone, speech recognition, Stanislav Petrov, stem cell, Steven Pinker, strong AI, synthetic biology, total factor productivity, transatlantic slave trade, Tyler Cowen, William MacAskill, women in the workforce, working-age population, World Values Survey, Y Combinator

So far, artificial intelligence has been narrow. AlphaGo is extraordinarily good at playing Go but is incapable of doing anything else.41 But some of the leading AI labs, such as DeepMind and OpenAI, have the explicit goal of building AGI.42 And there have been indications of progress, such as the performance of GPT-3, an AI language model which can perform a variety of tasks it was never explicitly trained to perform, such as translation or arithmetic.43 AlphaZero, a successor to AlphaGo, taught itself how to play not only Go but also chess and shogi, ultimately achieving world-class performance.44 About two years later, MuZero achieved the same feat despite initially not even knowing the rules of the game.45 The development of AGI would be of monumental longterm importance for two reasons.

…

More specifically, most AI breakthroughs have been due to a particular approach to machine learning that uses multilayered neural networks, known as “deep learning” (Goodfellow et al. 2016; LeCun et al. 2015). At the time of writing, the state-of-the-art AI for text-based applications are so-called transformers, which include Google’s BERT and OpenAI’s GPT-3 (T. Brown et al. 2020; Devlin et al. 2019; Vaswani et al. 2017). Transformers have also been successfully used for tasks involving audio (Child et al. 2019), images (M. Chen et al. 2020; Dosovitskiy et al. 2021), and video (Wang et al. 2021). The highest-profile AI achievements in real-time strategy games were DeepMind’s AlphaStar defeat of human grandmasters in the game StarCraft II and the OpenAI Five’s defeat of human world champions in Dota 2 (OpenAI et al. 2019; Vinyals et al. 2019).

Visual Thinking: The Hidden Gifts of People Who Think in Pictures, Patterns, and Abstractions by Temple Grandin, Ph.D.

2013 Report for America's Infrastructure - American Society of Civil Engineers - 19 March 2013, 3D printing, a long time ago in a galaxy far, far away, air gap, Albert Einstein, American Society of Civil Engineers: Report Card, Apollo 11, Apple II, ASML, Asperger Syndrome, autism spectrum disorder, autonomous vehicles, Black Lives Matter, Boeing 737 MAX, Captain Sullenberger Hudson, clean water, cloud computing, computer vision, Computing Machinery and Intelligence, coronavirus, cotton gin, COVID-19, defense in depth, Drosophila, Elon Musk, en.wikipedia.org, GPT-3, Gregor Mendel, Greta Thunberg, hallucination problem, helicopter parent, income inequality, industrial robot, invention of movable type, Isaac Newton, James Webb Space Telescope, John Nash: game theory, John von Neumann, Jony Ive, language acquisition, longitudinal study, Mark Zuckerberg, Mars Rover, meta-analysis, Neil Armstrong, neurotypical, pattern recognition, Peter Thiel, phenotype, ransomware, replication crisis, Report Card for America’s Infrastructure, Robert X Cringely, Saturday Night Live, self-driving car, seminal paper, Silicon Valley, Skinner box, space junk, stem cell, Stephen Hawking, Steve Jobs, Steve Wozniak, Tacoma Narrows Bridge, TaskRabbit, theory of mind, TikTok, twin studies, unpaid internship, upwardly mobile, US Airways Flight 1549, warehouse automation, warehouse robotics, web application, William Langewiesche, Y Combinator

In the journal Nature, David Silver and his colleagues write that the computer used “nonstandard strategies beyond the scope of traditional Go knowledge.” AI is being studied and applied in areas as diverse as video games and analyzing satellite images. AI programs are even being trained to write plays and essays. In an article on Medium.com, Sofia Merenych wrote about GPT-3, a program that composed a play so thoroughly in the manner of Shakespeare that linguists had a difficult time determining it was fake. When the program sucked up vast amounts of human knowledge off the internet, it was capable of writing essays on different subjects. When asked to write about a controversial subject, it sometimes came to conclusions that were offensive.

The Wires of War: Technology and the Global Struggle for Power by Jacob Helberg

"World Economic Forum" Davos, 2021 United States Capitol attack, A Declaration of the Independence of Cyberspace, active measures, Affordable Care Act / Obamacare, air gap, Airbnb, algorithmic management, augmented reality, autonomous vehicles, Berlin Wall, Bernie Sanders, Big Tech, bike sharing, Black Lives Matter, blockchain, Boris Johnson, Brexit referendum, cable laying ship, call centre, Cambridge Analytica, Cass Sunstein, cloud computing, coronavirus, COVID-19, creative destruction, crisis actor, data is the new oil, data science, decentralized internet, deep learning, deepfake, deglobalization, deindustrialization, Deng Xiaoping, deplatforming, digital nomad, disinformation, don't be evil, Donald Trump, dual-use technology, Edward Snowden, Elon Musk, en.wikipedia.org, end-to-end encryption, fail fast, fake news, Filter Bubble, Francis Fukuyama: the end of history, geopolitical risk, glass ceiling, global pandemic, global supply chain, Google bus, Google Chrome, GPT-3, green new deal, information security, Internet of things, Jeff Bezos, Jeffrey Epstein, John Markoff, John Perry Barlow, knowledge economy, Larry Ellison, lockdown, Loma Prieta earthquake, low earth orbit, low skilled workers, Lyft, manufacturing employment, Marc Andreessen, Mark Zuckerberg, Mary Meeker, Mikhail Gorbachev, military-industrial complex, Mohammed Bouazizi, move fast and break things, Nate Silver, natural language processing, Network effects, new economy, one-China policy, open economy, OpenAI, Parler "social media", Peter Thiel, QAnon, QR code, race to the bottom, Ralph Nader, RAND corporation, reshoring, ride hailing / ride sharing, Ronald Reagan, Russian election interference, Salesforce, Sam Altman, satellite internet, self-driving car, Sheryl Sandberg, side project, Silicon Valley, Silicon Valley ideology, Silicon Valley startup, Skype, smart grid, SoftBank, Solyndra, South China Sea, SpaceX Starlink, Steve Jobs, Steven Levy, Stuxnet, supply-chain attack, Susan Wojcicki, tech worker, techlash, technoutopianism, TikTok, Tim Cook: Apple, trade route, TSMC, Twitter Arab Spring, uber lyft, undersea cable, Unsafe at Any Speed, Valery Gerasimov, vertical integration, Wargames Reagan, Westphalian system, white picket fence, WikiLeaks, Y Combinator, zero-sum game

AI powers self-driving cars and suggests movies we might like on Netflix. The Associated Press has used AI to draft basic articles. IBM’s Watson beat two of Jeopardy!’s greatest contestants and, for good measure, identified genes linked to degenerative illness. In June 2020, the San Francisco company OpenAI’s GPT-3 sent shock waves across the tech industry, proving it possible to algorithmically generate cogent and naturally sounding long-form text on almost any topic. The consulting firm PwC estimates that artificial intelligence will contribute an additional $15.7 trillion to global economic growth by 2030.