large language model

The Infinity Machine: Demis Hassabis, DeepMind, and the Quest for Superintelligence

by Sebastian Mallaby;  · 30 Mar 2026  · 607pp  · 161,998 words

during DeepMind’s early days, but not after 2015 or so. The question of whether AI systems need to be “grounded” is still hotly debated. Large language models such as ChatGPT or Gemini are not directly taught concepts, yet these systems exhibit an impressive grasp of how the world functions. A feeling for

…

get through; Hassabis with his sparky riffs on intelligence and life, neuroscience and games, history and fiction. This was the period of maximum excitement about large language models, so language and how to think about it came up repeatedly in our sessions. At that crucial moment, when Ilya Sutskever had read the transformer

…

DeepMind’s agents could manage. Given the competition between reinforcement learning and deep learning, this setback held a warning. When it came to OpenAI’s large language models, the more complex and varied the data you fed into the system, the better it performed. Much as knowing French helps a linguist to master

…

Italian, large language models were capable of “transfer learning” across different but loosely related topics. But Gaia’s failure suggested that the same was not true for RL. A

…

, racing monomaniacally to build ever larger transformer models rather than daring to imagine more innovative approaches. Some of these attacks are baffling: The creators of large language models repeatedly roll out novel features, from text-and-video multimodality, to longer memory, to step-by-step reasoning; and the biases of the models have

…

Jack Rae had tried to drum up support for a rival DeepMind language project. But Hassabis and his lieutenants were skeptical of the potential of large language models, disinclined to follow OpenAI, and preoccupied with AlphaFold, StarCraft II, and governance wrangles with Google. Irving’s arrival tipped the balance at DeepMind. He had

…

, responsibility would require “stepping back to assess the situation we find ourselves in, mapping out potential risks, and researching mitigations.” The goal was to create “large language models that serve society, furthering our mission of solving intelligence to advance science and benefit humanity.”[22] The reassuring promises betrayed no hint of the reality

…

like Sam’s thing is, ‘I’m a pragmatic entrepreneur, I want to make amazing technology,’ ” Rae said later. “You go to OpenAI and really large language models are the bet. There’s no other bet. That’s very appealing.”[24] Irving could see Rae’s point—he understood the allure of a

…

arrived in 2021, Amodei had left to found Anthropic. Like Geoffrey Irving at DeepMind, Leike believed in pushing hard on the research but then releasing large language models cautiously. “Before we scramble to deeply integrate LLMs everywhere in the economy, can we pause and think whether it is wise to do so?” he

…

unknown, so the goal was to build a platform for exploratory research—Bell Labs had shown how you could do this. But the advent of large language models had scrambled the premise. The path ahead was now visible for all to see. The challenge was to set out on that path and to

…

, for two reasons. “One is that AI has got to the level where building a product is not going to divert me from building AGI. Large language models are both things: something you can sell, and something that advances the mission. “The second thing is that I’ve done product design before, and

…

dangerous outputs, such as advice on plotting an attack, could be removed by any half-sophisticated user. But LeCun swept this worry under the carpet. Large language models were not strong enough to make bad actors worse, he claimed. Besides, if the models were poised to become the fount of all knowledge, they

…

recently, other nations had joined the race. The governments of Saudi Arabia and the United Arab Emirates were planning to spend billions on their own large language models; in France, an ex-DeepMind scientist was joining forces with two Meta alumni to start a lab called Mistral. At a meeting with British Prime

…

difficulties of broader collaboration. By the summer of 2023, the Gemini project employed several hundred researchers. Their goal was to build the world’s strongest large language model by the end of the year, the deadline that the company had set for a GPT-4-level system. Scientists worked nights and weekends, often

…

an inspiring talk to his research colleagues, laying out how sophisticated, machine-based RL—going well beyond simple reinforcement learning from human feedback—could take large language models to the next level. But Silver quickly ran into a wall. It was partly that he was not cut out to work inside a product

…

Gemini we are matching that. And I think we still have the better ideas,” Hassabis added. I pressed a little harder. People increasingly spoke of large language models and ChatGPT as though they were interchangeable. It was like web search: People just called it “googling.” Valley venture capitalists believed that GPT’s brand

…

release, however, Hassabis believed that the cycle was about to turn again. “People say, ‘Oh, it used to be AlphaGo, now it’s all about large language models,’ ” Hassabis told me. “But this is just a moment in time. AlphaGo-type methods are coming back. These language systems are going to need planning

…

percent of them.[32] The system’s ability to ruminate—to backtrack and self-correct—was eerily human. After all, backtracking was antithetical to traditional large language models, which were trained to look forward. “They’re just kind of like, ‘predict next token, predict next token,’ ” an OpenAI researcher said, describing nonthinking models

…

a string of research triumphs, playing a lead role in the creation of the transformer architecture. “I have invented much of the current revolution in large language models,” Shazeer’s LinkedIn profile stated, and Google evidently agreed. Recently, after Shazeer had quit to launch an AI venture, the company had spent $2.7

…

had “just kind of vanished from the attention of the mainstream.” The field of artificial intelligence had descended into what Silver called the “valley” of large language models. The promise of RL had been forgotten. “You are the people who continued believing,” Silver told his audience. “At some point we need to get

…

we have to get past LLM Valley. “To become superhuman, the agent must interact and learn from its environment,” Silver went on. By mimicking humans, large language models had become impressively general. But the goal was to escape the echo chamber of existing knowledge—to uncover truths of which humans had no inkling

…

work of James Clerk Maxwell that people understood electricity, that there was a theory. I think deep learning is a bit similar. We can build large language models and they work very well, but nobody understands completely why. There is progress being made, but there is not yet a theory of it. I

…

TO NOTE REFERENCE 22 Silver even believed that a sufficiently large network could learn to search autonomously. He was anticipating the “emergent properties” that extremely large language models would exhibit several years later. Silver, author interview. BACK TO NOTE REFERENCE 23 Silver explains, “There are only two scalable things that we’ve really

…

-Shot Learners,” arXiv, February 8, 2022, doi.org/10.48550/arXiv.2109.01652. BACK TO NOTE REFERENCE 28 Jordan Hoffmann et al., “Training Compute-Optimal Large Language Models,” arXiv, March 29, 2022, doi.org/10.48550/arXiv.2203.15556. BACK TO NOTE REFERENCE 29 Sutskever, author interview. BACK TO NOTE REFERENCE 30 Amelia

…

, Empire of AI, 258. BACK TO NOTE REFERENCE 12 Will Douglas Heaven, “Why Meta’s Latest Large Language Model Survived Only Three Days Online,” MIT Technology Review, November 18, 2022, technologyreview.com/2022/11/18/1063487/meta-large-language-model-ai-only-survived-three-days-gpt-3-science. BACK TO NOTE REFERENCE 13 Sparrow’s search

…

It Is; A New Open-Source Security Threat: AIJacking.” BACK TO NOTE REFERENCE 4 Jason Wei et al., “Chain of Thought Prompting Elicits Reasoning in Large Language Models,” arXiv, January 28, 2022, arxiv.org/abs/2201.11903. Google’s release of this paper demonstrates that it was generally more open than OpenAI in

…

. BACK TO NOTE REFERENCE 21 Demis Hassabis, remarks at Imagination in Action (MIT), February 18, 2025. BACK TO NOTE REFERENCE 22 Jérémy Scheurer et al., “Large Language Models Can Strategically Deceive Their Users When Put Under Pressure,” arXiv, July 15, 2024, arxiv.org/abs/2311.07590. BACK TO NOTE REFERENCE 23 This behavior

…

, metr.org/blog/2025-06-05-recent-reward-hacking. BACK TO NOTE REFERENCE 25 Carson Denison et al., “Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models,” arXiv, June 14, 2024, arxiv.org/abs/2406.10162. BACK TO NOTE REFERENCE 26 Bowen Baker et al., “Monitoring Reasoning Models for Misbehavior and the

Prophecy: Prediction, Power, and the Fight for the Future, from Ancient Oracles to AI

by Carissa Véliz  · 21 Apr 2026  · 503pp  · 129,255 words

the probability that a given image contains a wolf, based on patterns it learned from thousands of images labeled wolf and not-wolf. When a large language model answers a question, it is predicting what a human being would say in its place, based on the statistical analysis of books, online forums, social

…

content produced by AI is akin to a designer drug in the form of a fortune cookie. Generative AIs Are Fortune Tellers, Not Truth Tellers Large language models like ChatGPT and other generative AI models are a good example both of how statistical predictions can have a tenuous relationship to truth and of

…

on the basis of pattern recognition through statistical analyses. Image generation models like DALL-E can create visual content by statistically processing millions of images. Large language models pick up language patterns by statistically analyzing millions of text samples. Chatbots develop a kind of map of how words, sentences, and concepts interconnect, allowing

…

response is one that a human being would be likely to give, and another human being likely to accept, not one that is probably true. Large language models are persuasive by design on two counts. First, they have analyzed enormous amounts of text, which allows them to mimic the patterns that they have

…

human beings’ tastes, thereby becoming more and more persuasive. But, as the proliferation of fake news has taught us, we don’t always prefer truth. Large language models are built to be fortune tellers, not truth tellers. And their persuasiveness makes them outstanding bullshitters (it’s a technical term, I promise—I’ll

…

else know far less than they proclaim. If Socrates was the wisest person in ancient Greece because he understood the limits of his knowledge, then large language models are foolish for the opposite reason: They don’t know what they don’t know. That also makes them the ultimate bullshitters. * * * — The philosopher Harry

…

Frankfurt argued that bullshit is speech that is typically persuasive but detached from a concern with the truth.[3] Large language models, as they are currently designed, are the ultimate bullshitters because they are made to be plausible with no regard for the truth. It probably didn

…

. L. Austin would put it, it’s necessary to investigate the speech situation (the context) to ascertain whether something is a prediction. * * * — All assertions by large language models are predictions, in the sense that they are a probabilistic output—even if they are assertions that are not about the future. Machine learning fills

…

is predicting what would be a plausible response from a human being, based on the texts and feedback it’s received. Say you ask a large language model to tell you what books Carissa Véliz has written, and it identifies me as the author of Privacy Is Power and Prophecy. That is true

…

a fact of the matter about whether I wrote a particular title, but does it count as knowledge if the answer was given by a large language model based on pure machine learning? One classic example can help us think through this matter. Imagine that you are visiting a city. While crossing the

…

get at the answer in a way that amounts to a claim that when read by a human justifies it being considered knowledge? If the large language model’s method were to corroborate titles and authors in a verified database, like the Library of Congress, then the answer would be an easy “yes

…

.” But that’s not what the large language model does; it calculates a likely response instead, which can include plausible but nonexistent titles, or it can miss one of the books I have written

…

human being. That’s why bullshit advice is dangerous: It sounds very much like good advice. The persuasive style but weak connection to truth of large language models makes them the perfect tool to create conspiracy theories and misinformation at scale. It also makes them dangerous to use in contexts like academia, medicine

…

least two elements out of three wrong: the paper’s title, first author, or year of publication.[13] Confabulation stems from the probabilistic structure of large language models. We would need to design language models differently to get rid of it. To put it another way, as long as models give responses based

…

of probability but of testimony. That’s causal thinking right there (and yes, also a prediction, smart aleck). We can already see it; responses from large language models have become better thanks to tech workers hand coding explicit guardrails into them (such as “don’t make up information”). Behind impressive

…

large language models lies a long list of system prompts guiding the models in non-probabilistic ways.[17] As Eamonn Maguire, director of AI at Proton, puts it,

…

. It’s supposed to ensure consistency and predictability in the application of the law. In 2023, a lawyer in New York City used ChatGPT, a large language model, to write up a brief for a case. The case was about a passenger who had sued an airline for getting injured during a flight

…

AI, but for all its novelty, from the point of view of privacy, it’s just another version of more of the same at scale. Large language models have ingested all the data on the internet available to them, including very personal data found in social media, forums, and databases of various kinds

…

could be cited on both sides of the Atlantic. BACK TO NOTE REFERENCE 35 Eubanks, Automating Inequality. BACK TO NOTE REFERENCE 36 Peters and Matz, “Large Language Models Can Infer Psychological Dispositions of Social Media Users”; Knight, “AI Chatbots Can Guess Your Personal Information.” For more on the ways in which tech is

…

Hard Time Raising Money from VCs.” Wired, Jan. 26, 2022. Parfit, Derek. On What Matters. Oxford: Oxford University Press, 2011. Peters, Heinrich, and Sandra Matz. “Large Language Models Can Infer Psychological Dispositions of Social Media Users.” PNAS Nexus 3, no. 6 (2024). Piña, Christy. “Mark Zuckerberg Clarifies Reports About His Bunker in Hawaii

…

Lover (Lawrence), 167 Laertius, Diogenes, 297 Langley, Samuel, 90 language translation, 69 Laplace, Pierre-Simon, 31, 37–40, 63, 65 Laplace’s demon, 65, 162 large language models, 93–94, 100–102, 106, 141 See also chatbots laurels, 295 Lavoisier, Antoine, 231 law, 107–8, 249, 290 Law, Steven J., 134 law. See

…

learning, 66, 68, 69, 91, 100, 103, 110, 119, 163, 180, 191, 219 causal inferences in, 105 in hybrid AI architecture, 106 images and, 103 large language models, 93–94, 100–102, 106, 141 oracles in, 68–69 surveillance and, 141 See also chatbots Macrinus, 117–18 Maduro, Nicolás, 50 Magi, 20 magic

…

, 38–40, 61–63, 65, 91, 102, 108, 109, 112, 240, 241 chatbots and, 93 frequentist definition of, 34, 36 insurance and, 48 interconnected, 115 large language models and, 101–2 productivity software, 71 Project Osmium, 80 prophecy. See prediction prophet’s dilemma, 121 Proton, 106 Prussia, 53, 249 pseudomysticism, 59 psychics, 122

We Are as Gods: A Survival Guide for the Age of Abundance

by Peter H. Diamandis and Steven Kotler  · 13 Apr 2026  · 225pp  · 76,418 words

create, converse, and even hallucinate. OpenAI got to work. After a year in the trenches, they unveiled GPT-1. Compared to contemporary standards, their first large language model was primitive at best. Yet it could take a prompt—a question, a headline, a half sentence—and spin out paragraphs that made sense. A

…

even forty. Try forty days. That’s about how long it took for social media to impact mental health. For AI to fake voices. For large language models to flood the web with synthetic noise. We’ve flipped the script. In Abundance, we focused on the upside of exponentials and the promise of

…

to connect those dots. That’s a staggering increase in lateral thinking. And here’s the point: Lateral thinking is a skill that AI lacks. Large language models are convergent problem-solvers that utilize deductive reasoning to match like with like. This lets AI excel at efficiency, speed, and raw computation, but wild

…

terms, centaurs combine human creative intuition and expertise with AI’s raw computational power. And with stunning results. In chess, for nearly a decade—before large language models (LLMs) rewrote the rules—centaurs outperformed both human grandmasters and solo machines. Kasparov’s experiment proved that, when harnessed correctly, AI can amp up human

…

experiments, Meta’s AI predicted words and phrases with up to 75 percent accuracy, suggesting the leap to brain-to-text translation is well underway. Large language models are accelerating this progress. LLMs trained on brain activity can now reconstruct sentences from fMRI scans with 40 percent accuracy. In a recent study, an

…

. Without it, our predictions falter. So feed curiosity on a regular basis. Ask great questions. Ask them often. This is one of the advantages of large language models. We all get to be young again, asking, “Why? Why? Why?” to our hearts’ content. Redefine Work as Creation: For a very long time, we

The Age of AI: And Our Human Future

by Henry A Kissinger, Eric Schmidt and Daniel Huttenlocher  · 2 Nov 2021  · 194pp  · 57,434 words

of AI are harder to quantify than increases in computing power, it appears that their growth is even more rapid. For example, the power of large language models, neural networks that underlie much of today’s natural language processing, is growing even more rapidly, tripling in fewer than two years. Microsoft’s Megatron

…

perhaps in which people play a de minimis co-inventor role). National governments have recognized AI’s threat to language: Hungary has commissioned its own large language model so that Hungarian does not automatically become obsolete in the digital realm.9 Governments have also begun to grapple with digital networks’ dilution of communal

Algospeak: How Social Media Is Transforming the Future of Language

by Adam Aleksic  · 15 Jul 2025  · 278pp  · 71,701 words

websites unusable.[3] That’s just from moving around tones, ignoring potential semantic substitutions, so we’re clearly quite far from a 1984-esque scenario. Large language models may get better at recognizing words in context, but people will always find creative ways to express their ideas. We’ve already seen this in

…

, 52–53 L labels, 196, 202, 205. See also microlabels Lady Bountiful (Carr), 18, 18 laggards, 53–55, 54 Language and the Internet (Crystal), 10n large language models, 209 late majority, 53, 54 Latin, 5, 7, 69n Latino dance circles, 147 laughing-crying emoji , 53–54, 188 laughing emoji, 53 LDAR (lay down

Literary Theory for Robots: How Computers Learned to Write

by Dennis Yi Tenen  · 6 Feb 2024  · 169pp  · 41,887 words

the frenzy of its grifters and soothsayers. What remains will be more modest and more significant. Viewed in the light of collective human intellectual achievement, large language models are built on the foundation of public archives, libraries, and encyclopedias containing the composite work of numerous authors. Their synthesized voice fascinates me not so

The Means of Prediction: How AI Really Works (And Who Benefits)

by Maximilian Kasy  · 15 Jan 2025  · 209pp  · 63,332 words

justice, and health care. In a similar vein, Timnit Gebru, a computer scientist writing during her time working at Google, warned of the dangers of large language models acting as stochastic parrots, which repeat language patterns without understanding, and in doing so replicate the biases embedded in their training data. Meredith Whittaker, currently

…

perceptions of AI have oscillated from “an obscure academic niche” to the very broad “everything related to data,” and back to the narrow category of “large language models.” This book will take an intermediate stance, in between these very broad and very narrow definitions. Following the standard technical treatment of AI, this book

…

measurable reward.” This definition is more specific than “anything to do with data,” but at the same time, it includes a lot more than just large language models. This definition gives us a framework to talk about the many socially consequential settings where AI is used. There are many branches of the field

…

outcome as accurately as possible. Many learning problems are prediction problems: In facial recognition, an individual’s identity is predicted based on an image. In large language models, the next word is predicted based on the preceding words. In the hiring of job candidates, future performance is predicted based on candidate characteristics. In

…

net), have also been central for generative AI—AI that produces text, images, or other media. This includes large language models, where the goal is to predict the most likely word to come next. (Large language models power applications such as ChatGPT.) Generative AI also includes image generation, where images are predicted based on text labels

…

. Public understanding has oscillated from AI being an obscure academic niche field, to it encompassing everything related to data, to it narrowly referring exclusively to large language models. This book will take an intermediate position, following leading textbooks on the subject. According to these textbooks, the goal of AI, as a field of

…

. A robot might choose which way to turn or how to move its arm. A social media algorithm might choose what ads to display. A large language model might choose what sentence to print next. A hiring algorithm might choose which job candidate to invite for an interview. There is, second, the objective

…

deadly set of applications of supervised learning has led to spectacular successes of AI in language processing, computer vision, and game play in recent years: Large language models can predict the word that is most likely to come next given a sequence of words typed thus far. Algorithms for machine translation predict the

…

of big data. Increasingly complex models, usually leveraging artificial neural nets, are fit on ever-larger datasets. (Examples of this development are the ever-larger large language models, which are discussed in the next chapter, where figure 6 shows the massive growth of both datasets and model complexity.) But no matter how large

…

still represent the same object. For natural language processing, transformer networks, another variant of neural networks that use self-attention, form the backbone of recent large language models. Such transformer networks are based on a recognition of the fact that, in human languages, the relevant context for interpreting a word often does not

…

the more general principles of AI and machine learning, which this book discusses, are likely to remain relevant in the long run. Today, virtually all large language models use the transformer architecture discussed above. The prime reason for the popularity of the transformer architecture is its practical success; its popularity is not motivated

…

6 shows the number of training tokens (i.e., the size of training datasets) and the number of parameters by year of release in different large language models. The vertical axis is scaled so that each step corresponds to a tenfold increase. Some of these models have more than one trillion parameters and

…

AI company DeepSeek suggest that this scale might be reduced somewhat without loss of performance.) Figure 6 The size of large language models Source: Wikipedia, “Large Language Model,” accessed October 1, 2024, https://en.wikipedia.org/wiki/Large_language_model Language modeling is one success story of deep learning; image generation is another one. The most successful algorithms for

…

approach is due to its practical success. Generative AI, whether for the generation of text or of images, raises important questions of data ownership. Both large language models and image generation models are trained on data that are primarily scraped from the internet, without either the knowledge or the consent of those who

…

maximizes or minimizes some objective function—such as predictive loss (i.e., prediction errors) for the next word on the internet, in the case of large language models. For generative AI, we need to modify this statement slightly: The model minimizes predictive loss for the next word in response to a prompt that

…

or benefits of their actions. Externalities might be positive or negative. Pollution is a negative externality, where others are harmed. The training and deployment of large language models, for example, consumes vast amounts of energy and thereby contributes to climate change. Knowledge production by contrast creates positive externalities, where others benefit. Public funding

…

an open-source license). These platforms greatly contribute to the common good. They also serve as one of the prime sources of training data for large language models, and for generative AI more broadly. All the works of art and writing that are available on the internet but that are considered someone’s

…

models. In a memorable and apt metaphor, the science fiction author Ted Chiang has described large language models, such as the one that underpins ChatGPT, as a blurry compressed image of the internet. The companies that control large language models are effectively scraping and compressing much of the internet, including all its contributions from a myriad

…

scalability, is the unstructured text data available on the internet. The comprehensive scraping of these data has powered much of the recent spectacular advances of large language models. Conceivably, however, limits might soon be reached in terms of the amount of text data of sufficient quality that can be made available. This suggests

…

alone do not make AI. To process data, we need computers. In application domains of AI where data are abundant, such as game play or large language models, the limiting factor for AI is usually computational capacity. The heart of a typical computer is its central processing unit, or CPU. The CPU is

…

and then setting the price equal to this value. Should the algorithm instead set prices to maximize consumer welfare? A chatbot algorithm based on a large language model might give answers corresponding to the most likely sequence of words that might follow a given question. The algorithm might do so by predicting responses

…

, 93 enclosures, 82, 108 energy: as component of the means of prediction, 93–95; consumption of, 93; externalities generated by production and consumption of, 95; large language models’ use of, 84; scalability of use of, 95 engineers/computer scientists: and AI ethics, 97–98; capacity of, as change agents, 100–101. See also

…

; competitive vs. noncompetitive, 150, 154; empirical trends in, 155–58; polarization of, 157; technology’s effects on, 149–50, 157, 161–62 Landemore, Hélène, 198 large language models: complexity of, 41; content of, 86, 87, 188; dangers of, 6; energy used by, 84; how they work, 10; as narrow definition of AI, 9

Artificial Intelligence: A Modern Approach

by Stuart Russell and Peter Norvig  · 14 Jul 2019  · 2,466pp  · 668,761 words

-speech tagger. In Proc. Sixth Conference on Applied Natural Language Processing. Brants, T., Popat, A. C., Xu, P., Och, F. J., and Dean, J. (2007). Large language models in machine translation. In EMNLP-CoNLL-07. Bratko, I. (2009). Prolog Programming for Artificial Intelligence (4th edition). Addison-Wesley. Bratman, M. E. (1987). Intention, Plans

The Singularity Is Nearer: When We Merge with AI

by Ray Kurzweil  · 25 Jun 2024

even experts expected—from mastering games like Jeopardy! and Go to driving automobiles, writing essays, passing bar exams, and diagnosing cancer. Now, powerful and flexible large language models like GPT-4 and Gemini can translate natural-language instructions into computer code—dramatically reducing the barrier between humans and machines. By the time you

…

survey found an aggregate prediction among AI experts that human-level machine intelligence would not arrive until around 2060.[11] But the latest advances in large language models have rapidly shifted expectations. As I was writing early drafts of this book, the consensus on Metaculus, the world’s top forecasting website, hovered between

…

“hairy body” or “trunk,” they often represent highly abstract statistical relationships that the model has discovered in its training data. Using these relationships, transformer-based large language models (LLMs) can predict which tokens would be most likely to follow a certain input prompt by a human. They then convert those back into text

…

’t just parroting a correct answer; it’s demonstrating deeper understanding by building up a coherent sequence of inferences step-by-step. Too often, though, large language models are so opaque that nobody can figure out how they arrived at a particular output. So by clarifying this process, PaLM both provides more trustworthy

…

exponentially. Recall the complexity ceiling idea from earlier in this chapter—similar math makes it very computation-intensive to increase the context window that a large language model can handle.[126] If there are ten wordlike ideas (i.e., tokens) in a given sentence, the number of possible relationships among subsets of them

…

-performance 2.7 times that of the TPU v4, measured on the MLPerf™ v3.1 Inference Closed benchmark, which is the gold standard for running large language models. To maximize commensurability with the TPU v4-4096 estimate, which approximates plausible high-volume contract pricing, TPU v5e pricing is estimated here per chip based

…

-research-using-ai-improve-chip-designs-2023-03-28. BACK TO NOTE REFERENCE 130 Blaise Aguera y Arcas, “Do Large Language Models Understand Us?,” Medium, December 16, 2021, https://medium.com/@blaisea/do-large-language-models-understand-us-6f881d6d8e75. BACK TO NOTE REFERENCE 131 With better algorithms, the amount of training compute needed to achieve a

…

on to work at Google, where he was a lead author of “Attention Is All You Need,” the paper that invented the transformer architecture for large language models that has powered the latest AI revolution. See Duke University, “Duke Researchers Pit Computer Against Human Crossword Puzzle Players,” ScienceDaily, April 20, 1999, https://www

…

GTX 680, 166, 310 gun culture, 230 guns, 3D-printed, 186–87 Gunsmoke (TV show), 144 Gutenberg, Johannes, 159, 160 H hallucinations, 65. See also large language models Hameroff, Stuart, 330n Hands Free Hectare, 202 Hanson, Robin, 60 Hanson Robotics, 101–2 hard problem of consciousness, 80–82, 92 hard takeoff, 60–61

…

argument, 48, 330n thought-to-text technology, 70–71 translation of, 48, 222 Turing test, 8–9, 9, 12–13, 60, 63–69, 71, 287 large language models (LLMs), 2, 13, 51, 55, 64–65 GPT-3, 47–48, 49, 52, 55, 239, 324n GPT-4, 2, 9, 52–56, 65 hallucinations, 65

…

, 255 O Oak Ridge National Laboratory, 61 observer selection bias, 98–99 occupational therapists, 198 On the Origin of Species (Darwin), 39 OpenAI. See also large language models ChatGPT, 52–53, 198 CLIP, 44 Codex, 50 DALL-E, 49–50 GPT-2, 47 GPT-3, 47–48, 49, 52, 55, 239, 324n GPT

AI in Museums: Reflections, Perspectives and Applications

by Sonja Thiel and Johannes C. Bernhardt  · 31 Dec 2023  · 321pp  · 113,564 words

a mainstream topic in the cultural world, but does feature in general debates about digitization and digitality. The use of machine learning, neural networks, and large language models has, however—and contrary to common assumptions—been growing for years. Beyond prominent lighthouses, initial surveys of the international museum landscape list many hundreds of

…

. The only thing that has changed dramatically in recent years is that such systems—from simple machine learning to the development of neural networks and large language models—have achieved a level of complexity and efficiency that often produces astonishing results. But to view this correctly, it is necessary to think the other

…

Creative User Empowerment. She places particular emphasis on reflecting the normative preconditions and frameworks of AI projects, stresses the importance of the conscious use of large language models and the open handling of data, and points to the requirement of clearly defining the problem to be solved with AI. Finally, the increasing use

…

art has played and continues to play a crucial role and might further transform and redesign creative processes. The increase in generative AI and especially large language models (LLMs) has led to a distortion of the public perception of what is meant by AI. At the same time, the technology has made astonishing

…

content, but also engaging in communicative interaction. It might be useful to remember in the future that there was a time before the development of large language models, and in ‘CHIM—Chatbot in the Museum: Exploring and Explaining Museum Objects with Speech-Based AI’, Oliver Guske, Stefan Schaffer, and Aaron Ruß present an

…

here is therefore rather particular: the model learns stochastically by adding up very small elements to calculate ‘a bigger picture’ or (in the case of large language models) to calculate the meaning of a sentence from analysing the context of thousands of tokens (entities similar to words), taking note of which other tokens

…

chatbot communication is now emerging as the technology develops and is implemented in our daily lives. As we know, this part of AI, which involves large language models and natural language processing, is not the only form of AI, but it is—besides generative images—nevertheless one that a majority of people currently

…

observations are based on a chatbot that is certainly sophisticated, but also far from exploiting the full potential of AI. With the rapid emergence of large language models (the best known being OpenAI’s GPT), museum chatbots will improve significantly over the next few years and be able to provide truly individualized responses

…

quantity of data and a model’s capacity to ingest this larger amount of information. This has been clearly illustrated with linguistic models (or LLMs, large language models). Behind the construction of larger models lies the idea of universality: by building larger models capable of sorting through a wider range of data, it

…

from the web—without any explanation or justification being provided. But not only image-based AI can suffer from bad quality or biased training data. Large language models like BERT, GPT, et cetera are also trained on massive 155 156 Part 2: Perspectives quantities of text that are scraped from publicly available online

…

was fun—the fun of make-believe. One could also say: the joy that comes from a well-told story. A fiction, not a fact. Large Language Models as Entertainment Is GPT-3 somewhat similar to the Mechanical Turk? No human intelligence whatsoever, but still capable of coming up with humanlike texts? Or

…

with you like a human, you react with human emotions. That was the case when machines spoke like very, very dumb humans. And now that large language models (LLM) can have long and coherent conversations with you, it becomes increasingly difficult to stay cold (Kilg 2021). So, is Anic really a decent column

…

columns that she has fallen in love with her own neural network? In short, are we supposed to believe such nonsense routinely made up by large language models? Counter-question: Why not? And, first of all: what exactly do we mean by ‘believe’ here? Do we believe all of Dostoevsky’s psychological aberrations

…

have to consciously give up our ‘disbelief’ in order to enjoy fiction. And we seem to enjoy doing this, in all sorts of different contexts. Large Language Models as Impostors One can assume that impostors exploit precisely this desire to let ourselves be deceived, this ‘willing suspension’. We tend to prefer grandiose and

…

to treating illnesses might turn out to be expensive, losing much of the healing magic on the way (Sacks 1990). But that is another story. Large Language Models as Storytellers—Used Best in which Contexts? The crucial question here: Would we trust a machine doctor in the first place? Surely knowledge retrieval as

…

image recognition and intelligent search technologies. On top of these novel approaches to the exploration of the collection, users are also invited to interact with large language models (LLMs) enriched with collection data, so that they can actively write texts about the objects and publicly share their story and findings with others. This

…

and further integrated into the development. In the actual tool, it supports the visual search process through image embeddings. An event on the use of large language models was held in July 2022 and a prototype developed, co-curated by the Turing Agency20 (Basel/Zurich/Berlin). This enabled us to explore the process

…

and the xCurator user journey. In the experimental Datalab, we were able to run various tests and develop solutions to test the added value of large language models (LLMs) for the xCurator solution. With this, it already became visible that LLMs can help users find and contextualize content and suggest topics, structures, and

…

. The use and application of generative AI with multimodal models falls within a broader ongoing debate surrounding large language models. Several AI researchers have issued an open call for a moratorium on the development of large language models such as ChatGPT or GPT for at least six months until further research on the technology has been

…

generative AI technologies produce can be factual, but might also be speculative. For this reason, generative text production as it occurs in the context of large language models such as ChatGPT or GPT-4 is often likened to the figure of the ‘stochastic parrot’ (Bender et al. 2021, 610–23): like a parrot

…

animals. Abstracts Mercedes Bunz, The Role of Culture in the Intelligence of AI Artificial intelligence has received a new boost from the recent hype about large language models. However, to avoid misconceptions, it is better to speak of ‘machine intelligence’. In addition to reflecting on current processes, the cultural sector can benefit from

…

uses its potential to cope with the flood of information. Daniel M. Feige, Why AI Cannot Think In the context of the recent interest in large language models (LLMs) and image creation using artificial intelligence, the debate about whether AI is capable of reasoning arises again and again. This paper argues that it

…

and information based on individual user interests, thus providing them with a personalized and more in-depth exploration of the collection. Users can interact with large language models (LLMs) enriched with collection data, thus enabling them to write about and share objects. Despite being experimental, this signifies a shift in the role of

Nexus: A Brief History of Information Networks From the Stone Age to AI

by Yuval Noah Harari  · 9 Sep 2024  · 566pp  · 169,013 words

More Everything Forever: AI Overlords, Space Empires, and Silicon Valley's Crusade to Control the Fate of Humanity

by Adam Becker  · 14 Jun 2025  · 381pp  · 119,533 words

In the Plex: How Google Thinks, Works, and Shapes Our Lives

by Steven Levy  · 12 Apr 2011  · 666pp  · 181,495 words

On the Edge: The Art of Risking Everything

by Nate Silver  · 12 Aug 2024  · 848pp  · 227,015 words

Superbloom: How Technologies of Connection Tear Us Apart

by Nicholas Carr  · 28 Jan 2025  · 231pp  · 85,135 words

Why Machines Learn: The Elegant Math Behind Modern AI

by Anil Ananthaswamy  · 15 Jul 2024  · 416pp  · 118,522 words

Stories Are Weapons: Psychological Warfare and the American Mind

by Annalee Newitz  · 3 Jun 2024  · 251pp  · 68,713 words

Supremacy: AI, ChatGPT, and the Race That Will Change the World

by Parmy Olson  · 284pp  · 96,087 words

Co-Intelligence: Living and Working With AI

by Ethan Mollick  · 2 Apr 2024  · 189pp  · 58,076 words

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

by Mustafa Suleyman  · 4 Sep 2023  · 444pp  · 117,770 words

The Long History of the Future: Why Tomorrow's Technology Still Isn't Here

by Nicole Kobie  · 3 Jul 2024  · 348pp  · 119,358 words

Code Dependent: Living in the Shadow of AI

by Madhumita Murgia  · 20 Mar 2024  · 336pp  · 91,806 words

Searches: Selfhood in the Digital Age

by Vauhini Vara  · 8 Apr 2025  · 301pp  · 105,209 words

Fixed: Why Personal Finance is Broken and How to Make it Work for Everyone

by John Y. Campbell and Tarun Ramadorai  · 25 Jul 2025

The Age of Extraction: How Tech Platforms Conquered the Economy and Threaten Our Future Prosperity

by Tim Wu  · 4 Nov 2025  · 246pp  · 65,143 words

Blank Space: A Cultural History of the Twenty-First Century

by W. David Marx  · 18 Nov 2025  · 642pp  · 142,332 words

Rise of the Robots: Technology and the Threat of a Jobless Future

by Martin Ford  · 4 May 2015  · 484pp  · 104,873 words

Four Battlegrounds

by Paul Scharre  · 18 Jan 2023

Pattern Breakers: Why Some Start-Ups Change the Future

by Mike Maples and Peter Ziebelman  · 8 Jul 2024  · 207pp  · 65,156 words

Boom: Bubbles and the End of Stagnation

by Byrne Hobart and Tobias Huber  · 29 Oct 2024  · 292pp  · 106,826 words

This Is for Everyone: The Captivating Memoir From the Inventor of the World Wide Web

by Tim Berners-Lee  · 8 Sep 2025  · 347pp  · 100,038 words

What If We Get It Right?: Visions of Climate Futures

by Ayana Elizabeth Johnson  · 17 Sep 2024  · 588pp  · 160,825 words

Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI

by Karen Hao  · 19 May 2025  · 660pp  · 179,531 words

These Strange New Minds: How AI Learned to Talk and What It Means

by Christopher Summerfield  · 11 Mar 2025  · 412pp  · 122,298 words

The Equality Machine: Harnessing Digital Technology for a Brighter, More Inclusive Future

by Orly Lobel  · 17 Oct 2022  · 370pp  · 112,809 words

The Sirens' Call: How Attention Became the World's Most Endangered Resource

by Chris Hayes  · 28 Jan 2025  · 359pp  · 100,761 words

Gilded Rage: Elon Musk and the Radicalization of Silicon Valley

by Jacob Silverman  · 9 Oct 2025  · 312pp  · 103,645 words

The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World

by Pedro Domingos  · 21 Sep 2015  · 396pp  · 117,149 words

Everything Is Predictable: How Bayesian Statistics Explain Our World

by Tom Chivers  · 6 May 2024  · 283pp  · 102,484 words

The Optimist: Sam Altman, OpenAI, and the Race to Invent the Future

by Keach Hagey  · 19 May 2025  · 439pp  · 125,379 words

The Thinking Machine: Jensen Huang, Nvidia, and the World's Most Coveted Microchip

by Stephen Witt  · 8 Apr 2025  · 260pp  · 82,629 words

The Wealth Ladder: Proven Strategies for Every Step of Your Financial Life

by Nick Maggiulli  · 22 Jul 2025

Amateurs!: How We Built Internet Culture and Why It Matters

by Joanna Walsh  · 22 Sep 2025  · 255pp  · 80,203 words

Unit X: How the Pentagon and Silicon Valley Are Transforming the Future of War

by Raj M. Shah and Christopher Kirchhoff  · 8 Jul 2024  · 272pp  · 103,638 words

Abundance

by Ezra Klein and Derek Thompson  · 18 Mar 2025  · 227pp  · 84,566 words

The Measure of Progress: Counting What Really Matters

by Diane Coyle  · 15 Apr 2025  · 321pp  · 112,477 words

If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All

by Eliezer Yudkowsky and Nate Soares  · 15 Sep 2025  · 215pp  · 64,699 words

Money in the Metaverse: Digital Assets, Online Identities, Spatial Computing and Why Virtual Worlds Mean Real Business

by David G. W. Birch and Victoria Richardson  · 28 Apr 2024  · 249pp  · 74,201 words

Enshittification: Why Everything Suddenly Got Worse and What to Do About It

by Cory Doctorow  · 6 Oct 2025  · 313pp  · 94,415 words

Extremely Hardcore: Inside Elon Musk's Twitter

by Zoë Schiffer  · 13 Feb 2024  · 343pp  · 92,693 words

Elon Musk

by Walter Isaacson  · 11 Sep 2023  · 562pp  · 201,502 words

The Big Fix: How Companies Capture Markets and Harm Canadians

by Denise Hearn and Vass Bednar  · 14 Oct 2024  · 175pp  · 46,192 words

The Origins of Efficiency

by Brian Potter  · 15 Feb 2025  · 474pp  · 134,246 words

The Economic Consequences of Mr Trump: What the Trade War Means for the World

by Philip Coggan  · 1 Jul 2025  · 96pp  · 36,083 words