machine translation

back to index

160 results

Machine Translation

by Thierry Poibeau  · 14 Sep 2017  · 174pp  · 56,405 words

the Modern Corporation, James W. Cortada Intellectual Property Strategy, John Palfrey The Internet of Things, Samuel Greengard Machine Learning: The New AI, Ethem Alpaydin Machine Translation, Thierry Poibeau Memes in Digital Culture, Limor Shifman Metadata, Jeffrey Pomerantz The Mind–Body Problem, Jonathan Westphal MOOCs, Jonathan Haber Neuroplasticity, Moheb Costandi Open

…

Self-Tracking, Gina Neff and Dawn Nafus Sustainability, Kent E. Portney The Technological Singularity, Murray Shanahan Understanding Beliefs, Nils J. Nilsson Waves, Frederic Raichlen Machine Translation Thierry Poibeau The MIT Press Cambridge, Massachusetts London, England © 2017 Massachusetts Institute of Technology All rights reserved. No part of this book may be reproduced

…

Report and Its Consequences 7 Parallel Corpora and Sentence Alignment 8 Example-Based Machine Translation 9 Statistical Machine Translation and Word Alignment 10 Segment-Based Machine Translation 11 Challenges and Limitations of Statistical Machine Translation 12 Deep Learning Machine Translation 13 The Evaluation of Machine Translation Systems 14 The Machine Translation Industry: Between Professional and Mass-Market Applications 15 Conclusion: The Future of

…

Machine Translation Glossary Bibliography and Further Reading Index About Author List of Tables Table 1 Example of possible

…

from the analysis of knowledge and reasoning, which explains the interest shown by philosophers and specialists of artificial intelligence as well as cognitive sciences in [machine translation]. Machine translation involves different processes that make it at least as challenging as developing an automatic dialoguing system. The degree of “understanding” shown by the machine

…

inferred information, which is of course highly challenging and simply goes beyond the current state of the art. The Revolution of Statistical Machine Translation Systems The classification of machine translation systems provided in the previous section is challenged by new approaches that have appeared since the early 1990s. The availability of huge quantities

…

in such symbols could be translated by all who possessed the dictionary” (Descartes, letter to Mersenne on November 20, 1629). This passage greatly inspired machine translation pioneers, since Descartes’ proposal aimed to replace words with unambiguous codes (“symbols” corresponding to numerical codes that are independent from the languages considered; symbols

…

in automatic translation systems. Although these projects resulted in relatively advanced proposals with vocabularies and grammar systems, they have rarely been actively used for machine translation. Esperanto was used during the 1980s in the European Distributed Translation Language project and within the Fujitsu company in Japan, but these two projects

…

it must therefore be translated; in other words, decoded in the target language). Beginning in 1947, Weaver corresponded with the cyberneticist Norbert Wiener concerning machine translation. He proposed that translation could be considered a “decoding” problem: One naturally wonders if the problem of translation could conceivably be treated as a problem

…

more powerful than symbolic rules to resolve ambiguities). The implementation of the proposed techniques, however, required efforts that went beyond anything the pioneers of machine translation had ever imagined. In particular, the inherent ambiguity of natural languages showed that traditional encryption models were not sufficient to render the complexity of automatic

…

-quality translation in the short or medium term (FAHQT, or fully automated high-quality translation; also found as FAHQMT, or fully automated high-quality machine translation). Instead of automatic translation, Bar-Hillel recommended that researchers turn toward computer-assisted translation systems, which constitute a relatively different project, clearly less exciting scientifically

…

the other hand, with the increasing amount of translations available on the Internet, it is now possible to directly design statistical models for machine translation. This approach, known as statistical machine translation, is the most popular today. Unlike a translation memory, which can be relatively small, automatic processing presumes the availability of an

…

it would be more convenient to directly use fragments of translation that one can find in existing bilingual corpora. An Overview of Example-Based Machine Translation Example-based machine translation typically operates in three stages to translate a given sentence: The system tries to find fragments of the sentence to be translated in the

…

sparsity (it is very difficult to collect enough relevant examples at phrase level). Appeal and Limitations of Example-Based Machine Translation Example-based machine translation generated great interest during the 1980s. Rather than developing a machine translation system manually, which is long and very costly, the example-based approach allowed for optimal exploitation of large

…

information of a syntactic and semantic nature has been progressively integrated into models to compensate for the limitation of purely statistical approaches. Toward Segment-Based Machine Translation The IBM models have been subjected to numerous enhancements. The most significant improvement was to take into consideration the notion of segments (or sequences

…

have described in this section have, however, helped improve the IBM models and can still be considered currently as the state of the art in machine translation. Introduction of Linguistic Information into Statistical Models Statistical translation models, despite their increasing complexity to better fit language specificities, have not solved all the

…

techniques. Lastly, for rare languages with too few data to make it possible to develop statistical systems, rule-based systems remain the norm. Hybrid Machine Translation Systems Following the success of statistical translation systems, the majority of traditional systems (based on large lexicons and transfer rules) gradually tried to incorporate statistical

…

deep learning provides an interesting approach that seems especially fitted for the challenges involved in improving human language processing. An Overview of Deep Learning for Machine Translation Deep learning achieved its first success in image recognition. Rather than using a group of predefined characteristics, deep learning generally operates from a very

…

to let the system infer by itself the best representation from the data. A translation system based solely on deep learning (aka “deep learning machine translation” or “neural machine translation”) thus simply consists of an “encoder” (the part of the system that analyzes the training data) and a “decoder” (the part of the

…

toward the resolution of such problems, hence the great success of this technique among researchers in the domain. Current Challenges for Deep Learning Machine Translation Until recently, machine translation systems based on deep learning performed well on simple sentences but still lagged behind traditional statistical systems for more complex sentences. There were different

…

of the internal model calculated by the neural network, so as to better understand how the whole approach works. The deep learning approach to machine translation (or neural machine translation) has proven efficient, first, on short sentences in closely related languages, and more recently on long sentences as well as more diverse languages.

…

initiated research in this area. A 1994 article (White et al., 1994) reviewed the first attempts at evaluation from the beginnings of research on machine translation. The article specifically reported the various possible strategies and their limits, described below. Comprehension Evaluation To assess comprehension, professional human translators first translated English newspaper

…

articles into different languages. Machine translation systems then translated the text back into English, and human analysts answered “multiple choice questions about the content of the articles” to evaluate the automatic

…

Fluency After the previous attempts involving human experts, DARPA then resorted to two evaluation scores: adequacy and fluency. As White and colleagues described of this machine translation (MT) evaluation method: “In an adequacy evaluation, literate, monolingual English speakers make judgments determining the degree to which the information in a professional translation

…

automatically. It is, rather, the capacity of the system to provide relevant translational elements that should be evaluated. 14 The Machine Translation Industry: Between Professional and Mass-Market Applications Machine translation is a popular application because it answers a very direct and simple need. Everybody can clearly see the importance of a system

…

analysis and translation systems used by Samsung’s connected devices (cell phones, tablets, and other technological gadgets). Facebook bought out different companies specialized in machine translation (such as Jibbigo in 2013 for voice messages in particular). Apple and Google are also regularly buying startups in the communication and information technology domains

…

is actually twofold. First, improving the productivity of translators: this involves efficient systems and strategies to make the best of the output of machine translation tools. Second, improving machine translation systems directly: this means being able to dynamically reuse end-user feedback to make the system evolve and propose more accurate translations in

…

Warren Weaver and the launching of MT: Brief biographical note”) and Y. Bar-Hillel (“Yehoshua Bar-Hillel: A philosopher’s contribution to machine translation”), both in Early Years in Machine Translation (see the full reference at the beginning of this chapter). Chapter 6: The 1966 ALPAC Report and Its Consequences The ALPAC report and

…

previous website (http://www.statmt.org) is probably the best source of information for recent trends related to statistical machine translation, of which segment-based machine translation is part. Chapter 11: Challenges and Limitations of Statistical Machine Translation See http://www.statmt.org,as for chapter 10 above. Kenneth Church (2011). “A pendulum swung too

…

Journal of Translation Studies 13 (1–2): 29–70. Special issue: The teaching of computer-aided translation, ed. Chan Sin Wai. Philipp Koehn (2009). Statistical Machine Translation. Cambridge: Cambridge University Press. Jorg Tiedemann (2011). Bitext Alignment. San Rafael, CA: Morgan and Claypool Publishers. Dan Jurafsky and James H. Martin (2016). Speech

…

3: The Correspondence. Cambridge: Cambridge University Press. Umberto Eco (1997). The Search for the Perfect Language. Oxford: Wiley. John Hutchins (2004). “Two precursors of machine translation: Artsrouni and Trojanskij.” International Journal of Translation 16 (1): 11–31. Philip P. Wiener (ed., 1951). Leibniz Selections. New York: Simon and Schuster. Yehoshua Bar

…

Human Intelligence (A. Elithorn and R. Banerji, eds.). Elsevier Science Publishers, Amsterdam. Eiichiro Sumita and Hitoshi Iida (1991). “Experiments and prospects of example-based machine translation.” Proceedings of the Twenty-Ninth Conference of the Association for Computational Linguistics, 185–192. Berkeley, CA. Thomas R. Green (1979). “The necessity of syntax markers

…

: Two experiments with artificial languages.” Verbal Learning and Verbal Behavior 18: 481–496. Harold Somers (1999). “Example-based machine translation.” Machine translation 14 (2): 113–157. Nano Gough and Andy Way (2004). “Robust large-scale EBMT with marker-based segmentation.” Proceedings of the Tenth International Conference on

…

Theoretical and Methodological Issues in Machine Translation, 95–104. Baltimore, MD. Peter Brown, John Cocke, Stephen Della Pietra, Vincent Della Pietra, Frederick Jelinek, Robert Mercer, and Paul Roossin (1988). “A

…

, Yoshua Bengio and Aaron Courville (2016). Deep Learning. Cambridge, MA: MIT Press. Yonghui Wu, et al. (2016). “Google's neural machine translation system: Bridging the gap between human and machine translation.” Published online. arXiv:1609.08144. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu (2002). “BLEU: A method for automatic

…

51, 56, 61, 63, 100, 118, 164–168, 215–216, 250 Environment Canada, 87, 223 Error rate, 241. See also Evaluation Errors (in machine translation). See Typology of errors in machine translation Escher, M. C., 20 Esperanto, 28, 42, 44 Estonian, 97, 212, 213 Europarl corpus, 97, 210, 223 Europe, 36, 41, 86, 97

…

58, 60, 85, 179 Logos Corporation, 88 Machine learning, 175, 181, 183, 236. See also Deep learning Machine translation evaluation. See Evaluation Machine translation industry. See Machine translation market Machine translation market, 89, 221–246, 247–251 Machine translation quality. See Evaluation Machine translation systems Apertium, 172 Ariane-78 system, 85 Babelfish, 227, 228 Bing Translation, 33, 36, 194, 226–229,

…

Systran) TAUM Météo (see Météo) Watson, 241 Maintenance applications, 243 Maltese, 212, 213 Manual correction, 138. See also Post-edition Mass-market applications of machine translation. See Machine translation market Mass media, 239 Mathematical model of communication. See Model of communication Meaning, 8, 15, 17–21, 34, 52–55, 64–67, 70–71

…

, Igor, 69 Memorandum, Warren Weaver’s, 50, 52–59 Mercer, Robert, 93, 94, 166, 216, 258 Mersel, Jules, 76 Metal. See Machine translation systems Metaphysics, 179 Météo. See Machine translation systems Meteor. See Evaluation measure and test Michigan University, 81 Microsoft, 227–229, 240, 248–250 MIT, 60–62 Mobile application, 229, 232

…

264 System comparison (see Evaluation) maintenance, 243 quality (see Evaluation) Systran, 85–89, 171, 194, 223, 226, 227–229, 231–236 Systranet. See Machine translation systems TAUM Météo. See Machine translation systems Technical text, 4, 11, 13, 88, 223, 244 Technical translation, 92. See also Technical text Terminology, 92, 101, 119, 226, 228, 244

Artificial Intelligence: A Guide for Thinking Humans

by Melanie Mitchell  · 14 Oct 2019  · 350pp  · 98,077 words

human language.” (In AI-speak, “natural” means “human.”) Natural-language processing (abbreviated NLP) includes topics such as speech recognition, web search, automated question answering, and machine translation. Similar to what we’ve seen in previous chapters, deep learning has been the driving force behind most of the recent advances in NLP. I

…

in AI, such “decoding” turned out to be harder than people originally expected. Like other AI research in the early days, the original approaches to machine translation relied on complicated sets of human-specified rules. With the goal of translating from a source language (for example, English) to a target language (for

…

instance, Russian), a machine-translation system would be given syntax rules for both languages as well as rules for mappings between syntactical structures. In addition, human programmers would create dictionaries

…

for the machine-translation system with word-to-word (and simple phrase-to-phrase) equivalences. Like many other efforts in symbolic AI, while these approaches worked well in some

…

Nations transcripts, which are translated into the six official languages of the UN, and from other large sets of original and translated documents. The statistical machine-translation systems of the 1990s to the 2000s typically computed large tables of probabilities linking phrases in the source and target languages. When given a new

…

worked together as a sentence, but the main driver of the translation was the probabilities of phrases learned from the training data. Even though statistical machine-translation systems had very little knowledge of syntax in either language, on the whole these methods produced better translations than the earlier rule-based approaches. Google

…

Translate—probably the most widely used automated-translation program—employed these kinds of statistical machine-translation methods from the time of its launch in 2006 until 2016, at which time Google researchers had developed what they claimed was a superior translation

…

method based on deep learning, called neural machine translation. Soon after, neural machine translation was adopted for all state-of-the-art machine-translation programs. Encoder, Meet Decoder Figure 38 gives a sketch of what’s under the hood when you use Google

…

neural network. While figure 38 shows the encoder and decoder networks abstractly as white rectangles, such networks are actually made up of LSTM units. Automated machine translation in the deep-learning age is a triumph of big data and fast computation. To create a pair of encoder-decoder networks to translate from

…

believe is that neural networks are learning the underlying semantic meaning of the language.”14 The CEO of the specialty translation company DeepL bragged, “Our [machine-translation] neural networks have developed an astounding sense of understanding.”15 In general, such declarations are in part fueled by the race among tech companies to

…

company and you want to translate a large volume of documents or provide translation for customers on your websites, you can find many fee-based machine-translation services available, all powered by the same encoder-decoder architecture. To what extent should we believe the claims that machines are actually learning “semantic meaning

…

” or that machine translation is swiftly closing in on human levels of accuracy? To answer this, let’s look more closely at the actual results these claims are based

…

to design an automatic method for computing the system’s accuracy. The claims of “human parity” and “bridging the gap between machines and humans” in machine translation are based on two methods of evaluating translation results. The first is an automated method—a computer program—that compares a machine’s translation with

…

out a score. The second method employs bilingual humans to manually evaluate translations. For the first method, the program used in virtually all evaluations of machine translation is called bilingual evaluation understudy, or BLEU.16 To measure the quality of a translation, BLEU essentially counts the number of matches—between words and

…

phrases of varying lengths—in a machine-translated sentence and one or more human-created “reference” (that is, “correct”) translations. While the ratings produced by BLEU often correlate with human judgments of translation

…

quality, BLEU tends to overrate bad translations. Several machine-translation researchers have told me that BLEU is a flawed way to evaluate translations, used only because no one has yet found an automatic method that

…

works better in general. Given the drawbacks of BLEU, the “gold standard” for evaluating a machine-translation system is for bilingual humans to manually rate the translations produced by the system. These same human evaluators can also rate corresponding translations created by

…

professional human translators in order to compare with the machine-translation ratings. But there are also drawbacks to this gold-standard approach: hiring humans costs money, of course, and unlike computers humans get tired after rating

…

you can hire an army of bilingual human raters who have a lot of time on their hands, your evaluation process will be limited. The machine-translation groups at both Google and Microsoft carried out this kind of gold-standard (albeit limited) evaluation by hiring small groups of bilingual human evaluators to

…

set of sentences in a source language, along with translations of those sentences into the target language. The translations were created both by the neural machine-translation system and by professional human translators. Google’s evaluation consisted of about five hundred sentences from news stories and from Wikipedia articles in several different

…

each evaluator’s ratings over all sentences, and then averaging over the evaluators, the Google researchers found that the average rating given to their neural machine-translation system was close to (though below) the ratings given to the human-translated sentences. This was the case for all of the language pairs in

…

. Microsoft used a similar averaging method to evaluate translations of news stories from Chinese to English. The ratings of the translations by Microsoft’s neural machine-translation system were very close to (and sometimes even exceeded) the ratings of the human translations. In all cases, the human evaluators rated the translations produced

…

one another in important ways that can be missed if the sentences are translated in isolation. I haven’t seen any formal studies of evaluating machine translation for longer passages, but my general experience is that the translation quality of, say, Google Translate declines significantly when it is given whole paragraphs instead

…

from news stories and Wikipedia pages, which are typically written with care to avoid ambiguous or idiomatic language; such language can cause serious problems for machine-translation systems. Lost in Translation Remember my “Restaurant” story from the beginning of the previous chapter? I didn’t design that story to test translation systems

…

, but the story actually does a good job of illustrating the challenges presented to machine-translation systems by colloquial, idiomatic, and potentially ambiguous language. I used Google Translate to translate the “Restaurant” story from English into three target languages: French, Italian

…

improved, and some of the specific translation errors seen here may be fixed by the time you are reading this. However, I’m skeptical that machine translation will actually reach the level of human translators—except perhaps in narrow circumstances—for a long time to come. The main obstacle is this: like

…

speech-recognition systems, machine-translation systems perform their task without actually understanding the text they are processing.21 In translation as well as in speech recognition, the question remains: To

…

for his meal than about “proposed legislation.” Hofstadter’s words were echoed in a recent article by the AI researchers Ernest Davis and Gary Marcus: “Machine translation … often involves problems of ambiguity that can only be resolved by achieving an actual understanding of the text—and bringing real-world knowledge to bear

…

is still an open question and is the subject of intense debate in the AI community. For now, I’ll simply say that while neural machine translation can be impressively effective and useful in many applications, the translations, without post-editing by knowledgeable humans, are still fundamentally unreliable. If you use

…

Symposium on Intelligent Data Analysis (2018), 328–39. 12: Translation as Encoding and Decoding   1.  Q. V. Le and M. Schuster, “A Neural Network for Machine Translation, at Production Scale,” AI Blog, Google, Sept. 27, 2016, ai.googleblog.com/2016/09/a-neural-network-for-machine.html.   2.  W. Weaver, “Translation,” in

…

Machine Translation of Languages, ed. W. N. Locke and A. D. Booth (New York: Technology Press and John Wiley & Sons, 1955), 15–23.   3.  This is the

…

for some less common languages.   4.  For more details, see Y. Wu et al., “Google’s Neural Machine Translation System: Bridging the Gap Between Human and Machine Translation,” arXiv:1609.08144 (2016).   5.  In Google’s neural machine-translation system, the word vectors are learned as part of the training of the entire network.   6.  More

…

network are probabilities for each possible word in the network’s vocabulary (here, French). More details are given in Wu et al., “Google’s Neural Machine Translation System.”   7.  At the time of this writing, Google Translate and other translation systems work by translating one sentence at a time. An example of

…

on going beyond sentence-by-sentence translation is described in L. M. Werlen and A. Popescu-Belis, “Using Coreference Links to Improve Spanish-to-English Machine Translation,” in Proceedings of the 2nd Workshop on Coreference Resolution Beyond OntoNotes (2017), 30–40.   8.  S. Hochreiter and J. Schmidhuber, “Long Short-Term Memory,” Neural

…

Computation 9, no. 8 (1997): 1735–80. 9. Wu et al., “Google’s Neural Machine Translation System.” 10.  Ibid. 11.  T. Simonite, “Google’s New Service Translates Languages Almost as Well as Humans Can,” Technology Review, Sept. 27, 2016, www.technologyreview

…

Historic Milestone, Using AI to Match Human Performance in Translating News from Chinese to English,” AI Blog, Microsoft, March 14, 2018, blogs.microsoft.com/ai/machine-translation-news-test-set-human-parity. 13.  “IBM Watson Is Now Fluent in Nine Languages (and Counting),” Wired, Oct. 6, 2016, www.wired.co.uk/article

…

of; inspiration from neuroscience; lack of reliability; as narrow AI; need for big data; see also convolutional neural networks; encoder-decoder system; encoder networks; neural machine translation; recurrent neural networks DeepMind; acquisition by Google; see also AlphaGo; Breakout deep neural networks, see deep learning deep Q-learning; adversarial examples for; on Breakout

…

Go (board game); see also AlphaGo Gödel, Escher, Bach (book) GOFAI Good, I. J. Goodfellow, Ian Google DeepMind, see DeepMind Google Translate; see also neural machine translation Gottschall, Jonathan GPS, see General Problem Solver GPUs, see graphical processing units gradient descent graphical processing units H HAL Hassabis, Demis Hawking, Stephen Hearst, Eliot

…

adversarial learning; bias in, see bias; interpretable, see explainable AI; overfitting in, see overfitting; transfer learning in, see transfer learning machine morality, see moral AI machine translation; comparison between humans and machines; evaluating; neural; statistical; see also Google Translate Manning, Christopher Marcus, Gary Markoff, John Marshall, James McCarthy, John McClelland, James Mechanical

…

: adversarial attacks on; challenges for; definition of; rule-based approaches to; statistical approaches to; see also machine translation; question answering; reading comprehension; sentiment classification; speech recognition; word vectors neocognitron network neural engineering neural machine translation; see also Google Translate; machine translation neural networks: activations in; classification in; convolutional, see convolutional neural networks; deep, see deep learning

…

, B. F. Smith, Brad speech recognition; adversarial examples for; word-error rate in Stanford Question Answering Dataset (SQuAD); human accuracy on Star Trek computer statistical machine translation strong AI; see also general or human-level AI subsymbolic AI; contrast with symbolic methods; integration with symbolic methods suitcase words Summer Vision Project (MIT

…

difference learning; see also reinforcement learning test set theory of mind thought vectors training, see supervised learning training set transfer learning; for Breakout translation, see machine translation trolley problem Turing, Alan Turing test; Kurzweil and Kapor wager on; Kurzweil’s predictions for U understanding: in analogy; ascribing to computers; in automated image

…

captioning; for creativity; in Cyc; in deep learning; in humans; in IBM Watson; in machine translation; for morality; for natural-language processing; in question-answering systems; for self-driving cars; in speech-recognition systems; in Star Trek computer; for vision; 263

Artificial Intelligence: A Modern Approach

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

, the MNIST data set for handwritten digit recognition, ImageNet and COCO for image object recognition, SQUAD for natural language question answering, the WMT competition for machine translation, and the International SAT Competitions for Boolean satisfiability solvers. AI was founded in part as a rebellion against the limitations of existing fields like control

…

learning systems have exceeded human performance on some vision tasks (and lag behind in some other tasks). Similar gains have been reported in speech recognition, machine translation, medical diagnosis, and game playing. The use of a deep network to represent the evaluation function contributed to ALPHAGO’S victories over the leading human

…

Google Maps provide driving directions for hundreds of millions of users, quickly plotting an optimal route taking into account current and predicted future traffic conditions. Machine translation: Online machine translation systems now enable the reading of documents in over 100 languages, including the native languages of over 99% of humans, and render hundreds of

…

for HMMs first appeared in Russell and Norvig (2003). HMMs have found many applications in language processing (Charniak, 1993), speech recognition (Rabiner and Juang, 1993), machine translation (Och and Ney, 2003), computational biology (Krogh et al., 1994; Baldi et al., 1994), financial economics (Bhar and Hamori, 2004) and other fields. There have

…

model that makes you $10 on every trade, that’s great—but not if it costs you $20 in computation cost for each prediction. A machine translation program that runs on your phone and allows you to read signs in a foreign city is helpful—but not if it runs down the

…

computation paths from inputs to outputs have many steps. Deep learning is currently the most widely used approach for applications such as visual object recognition, machine translation, speech recognition, speech synthesis, and image synthesis; it also plays a significant role in reinforcement learning applications (see Chapter 23). Deep learning has its origins

…

algorithm for each experiment. This has encouraged an approach called end-to-end learning, in which a complex computational system for a task such as machine translation can be composed from several trainable subsystems; the entire system is then trained in an end-to-end fashion from input/output pairs. With this

…

translation tasks. Supervised translation consists of gathering many (x, y) pairs and training the model to map each x to the corresponding y. For example, machine translation systems are often trained on pairs of sentences that have been translated by professional human translators. For other kinds of translation, supervised training data may

…

with near-perfect accuracy. 22.8.2Natural language processing Deep learning has also had a huge impact on natural language processing (NLP) applications such as machine translation and speech recognition. Some advantages of deep learning for these applications include the possibility of end-to-end learning, the automatic generation of internal representations

…

encoders and decoders. End-to-end learning refers to the construction of entire systems as a single, learned function f. For example, an f for machine translation might take as input an English sentence SE and produce an equivalent Japanese sentence SJ = f (SE). Such an f can be learned from training

…

. (2016b) showed that end-to-end translation using deep learning reduced translation errors by 60% relative to a previous pipeline-based system. As of 2020, machine translation systems are approaching human performance for language pairs such as French and English for which very large paired data sets are available, and they are

…

for image processing and other tasks where the data have a grid topology. •Recurrent networks are effective for sequence-processing tasks including language modeling and machine translation. Bibliographical and Historical Notes The literature on neural networks is vast. Cowan and Sharp (1988b, 1988a) survey the early history, beginning with the work of

…

about 2/3 of listeners saying that the neural WaveNet system (van den Oord et al., 2016a) sounded more natural than the previous nonneural system. Machine translation transforms text in one language to another. Systems are usually trained using a bilingual corpus: a set of paired documents, where one member of the

…

pair is in, say, English, and the other is in, say, French. The documents do not need to be annotated in any way; the machine translation system learns to align sentences and phrases and then when presented with a novel sentence in one language, can generate a translation to the other

…

covers the use of recurrent neural networks to capture meaning and long-distance context as text is processed sequentially. Section 25.3 focuses primarily on machine translation, one of the major successes of deep learning applied to NLP. Sections 25.4 and 25.5 cover models that can be trained from large

…

sentences with many intervening words between the subject and verb. 25.3Sequence-to-Sequence Models One of the most widely studied tasks in NLP is machine translation (MT), where the goal is to translate a sentence from a source language to a target language—for example, from Spanish to English. We train

…

a basic sequence-to-sequence model, an example of which is shown in Figure 25.6. Sequence-to-sequence models are most commonly used for machine translation, but can also be used for a number of other tasks, like automatically generating a text caption from an image, or summarization: rewriting a long

…

the target RNN. Unlike most components of neural networks, attention probabilities are often interpretable by humans and intuitively meaningful. For example, in the case of machine translation, the attention probabilities often correspond to the word-to-word alignments that a human would generate. This is shown in Figure 25.7(b). Sequence

…

-to-sequence models are a natural for machine translation, but almost any natural language task can be encoded as a sequence-to-sequence problem. For example, a question-answering system can be trained on

…

called the transformer encoder. It is useful for text classification tasks. The full transformer architecture was originally designed as a sequence-to-sequence model for machine translation. Therefore, in addition to the encoder, it also includes a transformer decoder. The encoder and decoder are nearly identical, except that the decoder uses a

…

be used to train a question-answering system. Similarly, many Web sites publish side-by-side translations of texts, which can be used to train machine translation systems. Some text even comes with labels of a sort, such as review sites where users annotate their text reviews with a 5-star rating

…

. If this model is trained on a large corpus of text, it generates pretrained representations that perform well across a wide variety of NLP tasks (machine translation, question answering, summarization, grammaticality judgments, and others). 25.6State of the art Deep learning and transfer learning have markedly advanced the state of the art

…

can effectively model local and long-distance context by retaining relevant information in their hidden-state vectors. •Sequence-to-sequence models can be used for machine translation and text generation problems. •Transformer models use self-attention and can model long-distance context as well as local context. They can make effective use

…

) while maintaining high accuracy. The XLM system (Lample and Conneau, 2019) is a transformer model with training data from multiple languages. This is useful for machine translation, but also provides more robust representations for monolingual tasks. Two other important systems, GPT-2 (Radford et al., 2019) and T5 (Raffel et al., 2019

…

France the same time zone as the UK?”) and two reading comprehension tasks involving answering questions after reading either a paragraph or a news article. Machine translation is a major application of language models. In 1933, Petr Troyanskii received a patent for a “translating machine,” but there were no computers available to

…

computing resources to make the approach practical. In the 1970s that began to change, and the SYSTRAN system (Toma, 1977) was the first commercially successful machine translation system. SYSTRAN relied on lexical and grammatical rules hand-crafted by linguists as well as on training data. In the 1980s, the community embraced purely

…

but not fluent results (Och and Ney, 2004; Zollmann et al., 2008). Och and Ney (2002) show how discriminative training led to an advance in machine translation in the early 2000s. Sutskever et al. (2015) first showed that it is possible to learn an end-to-end sequence-to-sequence neural model

…

for machine translation. Bahdanau et al. (2015) demonstrated the advantage of a model that jointly learns to align sentences in the source and target language and to translate

…

between the languages. Vaswani et al. (2018) showed that neural machine translation systems can further be improved by replacing LSTMS with transformer architectures, which use the attention mechanism to capture context. These neural translation systems quickly overtook

…

workers face, and free them to concentrate on more interesting aspects. People with disabilities will benefit from AI-based assistance in seeing, hearing, and mobility. Machine translation already allows people from different cultures to communicate. Software-based AI solutions have near zero marginal cost of production, and so have the potential to

…

, IBM, Salesforce) have begun competing to offer machine learning APIs with pre-built models for specific tasks such as visual object recognition, speech recognition, and machine translation. These models can be used as is, or can serve as a baseline to be customized with your particular data for your particular application. We

…

adversarial networks (GANs) and transformer language models each opened up new areas of research. We have also seen steps towards “diversity of behaviour.” For example, machine translation systems in the 1990s were built one at a time for each language pair (such as French to English), but today a single system can

…

. and Schneider, J. (2001). Autonomous helicopter control using reinforcement learning policy search methods. In ICRA-01. Bahdanau, D., Cho, K., and Bengio, Y. (2015). Neural machine translation by jointly learning to align and translate. In ICLR-15. Bahubalendruni, M. R. and Biswal, B. B. (2016). A review on assembly sequence generation and

…

. 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, and Practical Reason

…

, G., Jones, L., Parmar, N., Schuster, M., Chen, Z., Wu, Y., and Hughes, M. (2018). The best of both worlds: Combining recent advances in neural machine translation. In ACL-18. Chen, S. F. and Goodman, J. (1996). An empirical study of smoothing techniques for language modeling. In ACL-96. Chen, T. and

…

planning. In ECML-06. Koditschek, D. (1987). Exact robot navigation by means of potential functions: Some topological considerations. In ICRA-87. Koehn, P. (2009). Statistical Machine Translation. Cambridge University Press. Koelsch, S. and Siebel, W. A. (2005). Towards a neural basis of music perception. Trends in Cognitive Sciences, 9, 578–584. Koenderink

…

systematic comparison of various statistical alignment models. Computational Linguistics, 29, 19–51. Och, F. J. and Ney, H. (2004). The alignment template approach to statistical machine translation. Computational Linguistics, 30, 417–449. Och, F. J. and Ney, H. (2002). Discriminative training and maximum entropy models for statistical

…

machine translation. In COLING–02. Ogawa, S., Lee, T.-M., Kay, A. R., and Tank, D. W. (1990). Brain magnetic resonance imaging with contrast dependent on blood

…

, D. M., Smith, A. F. M., and Makov, U. E. (1985). Statistical Analysis of Finite Mixture Distributions. Wiley. Toma, P. (1977). SYSTRAN as a multilingual machine translation system. In Proc. Third European Congress on Information Systems and Networks: Overcoming the Language Barrier. Tomasi, C. and Kanade, T. (1992). Shape and motion from

…

., Le, Q. V., Norouzi, M., Macherey, W., Krikun, M., Cao, Y., Gao, Q., Macherey, K., et al. (2016b). Google’s neural machine translation system: Bridging the gap between human and machine translation. arXiv:1609.08144. Wu, Y. and He, K. (2018). Group normalization. arXiv:1803.08494. Xiong, W., Wu, L., Alleva, F., Droppo, J

…

Experimentation by Computer. MIT Press. Yngve, V. (1955). A model and an hypothesis for language structure. In Locke, W. N. and Booth, A. D. (Eds.), Machine Translation of Languages. MIT Press. Yob, G. (1975). Hunt the wumpus! Creative Computing, Sep/Oct. Yoshikawa, T. (1990). FoundationsofRobotics: Analysis and Control. MIT Press. You, Y

…

, 1117 Macherey, W., 47, 834, 901, 916, 1090, 1117 Machina, M., 550, 1105 machine evolution, 39 machine learning, 20, 22, 53, 220, 669, 669–987 machine translation (MT), 900, 915, 931 unsupervised, 831 Machover, M., 297, 1087 Macià, N., 737, 1097 MacKay, D. J. C., 476, 738, 1105, 1106 MacKenzie, D., 331

…

predictive control), 963, 971 MPE, see explanation, most probable MPI (mutual preferential independence), 533 MRS (metalevel reasoning system), 316 MST (minimum spanning tree), 127 MT (machine translation), 900, 915, 931 MUC (Message Understanding Conference), 905 Mudd, H. R, 1058 Mudigonda, M., 48, 1103 Mueller, E. T., 334, 358, 1107, 1113 Muggleton, S

…

language, 884 of probability, 410 synthesis (by theorem provers), 327 of algorithms, 327 SYNTHIA (simulated environment), 873 Syrjänen, T., 360, 1107 systematic search, 93 SYSTRAN (machine translation software), 931 Syverson, C., 1049, 1089 Szafron, D., 637, 638, 1087, 1088, 1116 Szathmáry, E., 161, 1113 Szegedy, C., 327, 330, 837, 838, 1085, 1086

…

kernel, 462 transition model, 70, 83, 123, 140, 193, 256, 479, 482, 510, 514, 553, 585, 939 sparse, 560 transitivity (of preferences), 520 translation, see machine translation transparency transpose, 1077 transposition (in a game), 201 transposition table, 201 Trappenberg, T., 839, 1115 traveling salesperson problem (TSP), 88, 125, 127 Traverso, P., 399

The Technology Trap: Capital, Labor, and Power in the Age of Automation

by Carl Benedikt Frey  · 17 Jun 2019  · 626pp  · 167,836 words

judge the accuracy of algorithmic translators in reproducing observed human translations. Every United Nations report, which is always translated by humans into six languages, gives machine translators more examples to learn from.10 And as the supply of data expands, computers do better. Google Translate draws on a plethora of algorithms, but

…

, but the networks performed poorly due to constraints imposed by computational resources. So up until recently, machine translations relied on algorithms that analyzed phrases word by word from millions of human translations. However, phrase-based machine translations suffered from some serious shortcomings. In particular, the narrow focus meant that the algorithm often lost the

…

this problem has been found in so-called deep learning, which uses artificial neural networks with more layers. These advances allow machine translators to better capture the structure of complex sentences. Neural Machine Translation (NMT), as it is called, used to be computationally expensive both in training and in translation inference. But due to

…

the progression of Moore’s Law and the availability of larger data sets, NMT has now become viable. In machine translation, deep learning is not without its own drawbacks. One major challenge relates to the translation of rare words. For example, if you type the Japanese

…

consistent with the general observation that AI is being applied to a cascading variety of tasks. The same technology that has shown promising results in machine translation is also performing visual tasks, such as image recognition. Starting from the individual pixels in an image, these algorithms work up through increasingly complex features

…

. A. Tanner, 2007. “Google Seeks World of Instant Translations,” Reuters, March 27. 11. Y. Wu et al., 2016, “Google’s Neural Machine Translation System: Bridging the Gap between Human and Machine Translation,” preprint, submitted October 8, https://arxiv.org/pdf/1609.08144.pdf. 12. I. M. Cockburn, R. Henderson, and S. Stern, 2018, “The

…

. Wu, Y., M. Schuster, Z. Chen, Q. V. Le, M. Norouzi, W. Macherey, M. Krikun, et al. 2016. “Google’s Neural Machine Translation System: Bridging the Gap between Human and Machine Translation.” Preprint, submitted September 26. https://arxiv.org/abs/1609.08144. Xiong, W., L. Wu, F. Alleva, J. Droppo, X. Huang, and A

…

, 303; JD. com, 313; Kiva Systems, 311; machine social intelligence, 317; Microsoft, 306; misconception, 311; multipurpose robots, 327; Neural Machine Translation, 304; neural networks, 303, 305, 314; pattern recognition, 319; phrase-based machine translation, 304; Siri (Apple), 306; speech recognition technology, 306; Turing test, 317; virtual agents, 306; voice assistant, 306; warehouse automation, 314

…

Administration, 178 nation states, rise of, 57 Nazi Labor front, 12 necessity, technological advances emerging from, 76 Neolithic communities, 34 Neolithic revolution, 33, 61 Neural Machine Translation (NMT), 304 neural networks, 303, 305, 314 Newcomen, Thomas, 53, 106, 317 New Deal, 200, 212, 272, 325 Newton, Isaac, 54 New World, discovery of

Surfaces and Essences

by Douglas Hofstadter and Emmanuel Sander  · 10 Sep 2012  · 1,079pp  · 321,718 words

the recycling bin. Given all this, how can we explain the fact that, in terms of serious thought, machines lag woefully behind us? Why is machine translation so often inept and awkward? Why are robots so primitive? Why is computer vision restricted to the simplest kinds of tasks? Why is it that

…

in English, but it has been coded in some strange symbols. I will now proceed to decode.’ ” Weaver’s humorous statement expresses the credo underlying machine translation, which is that translation is an act of “decoding” essentially analogous to using a substitution cipher, in which, in order to encode or decode a

…

by its alphabetic predecessor, in this case yielding Abraham Lincoln’s famous opening gambit, “Four score and seven years ago”. In the early days of machine translation, translation between languages was seen as this same process but just on a larger scale, in the sense that it operated not on letters but

…

Language A. We might test the efficacy of this strategy by seeing how the world’s most readily available (and perhaps also its most sophisticated) machine-translation “engine” performs on our little Lincolnian phrase “Four score and seven years ago”. (The term “engine” is a friendly tip of the hat to Charles

…

to come up with appropriate analogues in Language B for strings of characters written in Language A. Only a few years after the dream of machine translation was hatched, it was already starting to run into profound problems. These problems were articulated by a number of skeptics, of whom perhaps the most

…

recreate some of the high-sounding flavor of Abraham Lincoln’s immortal phrase, while sidestepping various superficially enticing traps along the way. Potential Progress in Machine Translation The preceding anecdote confirms the pervasive thesis of Warren Weaver’s book Alice in Many Tongues, which is that to translate well, the use of

…

and then to judge their appropriateness, one must carefully exploit one’s full inventory of mental resources, including one’s storehouse of life experiences. Could machine translation possibly do anything of the sort? Is it conceivable that one day, computer programs will be able to carry out translation at a high level

…

of artistry? A couple of decades ago, some machine-translation researchers, spurred by the low quality of what had then been achieved in their field, began to question the methods on which the field had

…

. What emerged with considerable vigor was the idea of statistical translation, which today has become a very important strategy used in tackling the challenge of machine translation. This approach is based on the use of statistically-based educated guesswork, where the data base in which all guesses are rooted consists of an

…

a data base is a marvelous treasurehouse of linguistic information, if only one can figure out how to exploit it. The basic idea of statistical machine translation is to choose among the many possible meanings of a “chunk” (that is, a word or several-word segment) in a piece of input text

…

order to do so, we’ll take a careful look at a short piece of French text in order to see how two extremely different machine-translation programs dealt with it — one using the old strategy, and one using the new strategy. The passage we will examine is taken from an obituary

…

the translation furnished by Google’s translation engine shortly after the obituary appeared. At that time, the Google engine was based on the original “Weaverian” machine-translation philosophy — namely, first via lookup in a very big on-line dictionary, followed by enhancement using grammatical “patching”. Original paragraph from Le Monde, September 2004

…

as they is merry of narguer the fate. It is obvious that the “decoding” technique — the technique that lay behind the original optimistic vision of machine translation — was hopelessly inadequate to the task, since the output that the translation engine yielded is pretty much nonsensical to an English speaker. It is ironic

…

apparently not in the engine’s on-line dictionary, so it was simply left in French.) This example gives a sense for the quality of machine translation in the fall of 2004. But now let us fast-forward to the spring of the year 2009. At that point, Google’s translation-engine

…

developers had radically switched strategies in favor of the new idea of statistical machine translation, so their new engine had little in common, other than its name, with its former incarnation. Given the inadequacy of the old method, which we

…

have any idea of the glee she took in laughing in the face of destiny. On reading this hand-done translation along with the two machine translations given earlier, one sees that humans and translation engines are not playing on the same turf — in fact, they are not even playing at the

…

languages no less than in monolingual speech production. But the analogies involved in carrying out human translation are not of the sort exploited by current machine-translation techniques, which make surface-level analogies between linguistic “chunks” based on statistical information that can be extracted, using intense calculation, from human-translated bilingual data

…

to make a large, coherent structure. Various and Sundry Challenges in Translating this Book The preceding section’s purpose was not to deride efforts at machine translation. Indeed, the blinding speed with which virtually anyone today can, entirely for free, get a glimpse of what is going on in a piece of

…

it to the New York subway, it would have been a wooden, almost mechanical kind of translation (although light-years more sophisticated than today’s machine translation). Could we not have elected, after having chosen an airport scene over a subway scene in the American version, to go back to the original

…

Cognition, 23, pp. 946–967. Khong, Yuen Foong (1992). Analogies at War. Princeton: Princeton University Press. Locke, W. N. and A. D. Booth (eds.) (1955). Machine Translation of Languages. Cambridge, Mass.: MIT Press. Macrae, C. Neil, Charles Stangor, and Miles Hewstone (1996). Stereotypes and Stereotyping. New York: The Guilford Press. Martin, Shirley

…

(1990). “Analog retrieval by constraint satisfaction”. Artificial Intelligence, 46, pp. 259–310. Weaver, Warren (1955). “Translation”. In W. N. Locke and A. D. Booth (eds.), Machine Translation of Languages. Cambridge, Mass.: MIT Press, pp. 15–23. ————— (1964). Alice in Many Tongues. Madison, Wisconsin: University of Wisconsin Press. Wharton, Charles M., Keith J

…

, list of, 136; stereotypes of, 135–136, 392, 486; strength of, as reflecting number of resemblances, 516; as strokes of genius, 16–17; superficial, in machine translation, 373, 375; taboo cases of, 104; as training wheels, 392; trivial and meaningless, ceaseless production of, 282, 284–286; unconscious, 259–281, 282, 285–286

…

unconscious analogies, 156 Bezout, Étienne, 413, 415, 420 bibles, category of 220, 229 bike-rental anomaly, explained by analogy, 328–330 bilingual data bases in machine translation, 369, 372–373 biplans: involving actions, 279; linguistic, 268–270 bird: as an example of an imprecise category, 55–56, 58, 59–60; as a

…

–35, 44 “lustre”, as possible French analogue to English word “score” in Gettysburg-address translation challenge, 372 —M— Macbeth effect, 289–290 Mach, Ernst, 487 machine translation, see translation Madonna, 223; syllogistic proof of mortality of, 193 magical angel stung by randomly buzzing interplanetary bumblebee, 493 magnet in motion, giving rise to

…

of, 442–443 squares, as questionable rectangles, 234–238, 255 staircases, negotiated by analogy, 507, 509, 516 Stargell, Willie, 325–326, 383 statistical approach to machine translation, 372–374 staying on the surface versus going into depth, 344 stealing, conceptual halo around, 106–107 stereotypes: of analogy-making, 135–136, 392, 521

…

, 386, 390, 403–407; see also categorical blinders, errors, fleeting analogies, latent analogies, naïve analogies understanding: as becoming used to something, 416; bypassing of, in machine translation, 368–370, 372–375; indispensability of, in human translation, 375–377 undo, on-line concept of, frame-blended with physical world, 406 undressing a banana

Natural language processing with Python

by Steven Bird, Ewan Klein and Edward Loper  · 15 Dec 2009  · 504pp  · 89,238 words

widespread. For example, phones and handheld computers support predictive text and handwriting recognition; web search engines give access to information locked up in unstructured text; machine translation allows us to retrieve texts written in Chinese and read them in Spanish. By providing more natural human-machine interfaces, and more sophisticated access to

…

solve such problems of language understanding, we will be able to move on to tasks that involve generating language output, such as question answering and machine translation. In the first case, a machine should be able to answer a user’s questions relating to collection of texts: (5) a. Text: ... The thieves

…

pronoun are steps in establishing the meaning of a sentence, things we would expect a language understanding system to be able to do. Machine Translation For a long time now, machine translation (MT) has been the holy grail of language understanding, ultimately seeking to provide high-quality, idiomatic translation between any pair of languages

…

0> The pig that John found looked happy 1> Das Schwein, das John fand, schaute gl?cklich 2> The pig, which found John, looked happy Machine translation is difficult because a given word could have several possible translations (depending on its meaning), and because word order must be changed in keeping with

…

.10 Exercises 1. ○ Read up on one of the language technologies mentioned in this section, such as word sense disambiguation, semantic role labeling, question answering, machine translation, or named entity recognition. Find out what type and quantity of annotated data is required for developing such systems. Why do you think a large

…

decision trees, 242–245 Maximum Entropy classifiers, 251–254 naive Bayes classifiers, 246–250 packages, 237 resources for further reading, 257 supervised classification, 221–237 machine translation (MT) limitations of, 30 using NLTK’s babelizer, 30 mapping, 189 Matplotlib package, 168–170 maximal projection, 347 Maximum Entropy classifiers, 251–254 Maximum Entropy

…

CoNLL 2000 Chunking Corpus, 273 training sets, 223, 225 transformation-based tagging, 208–210 transitive verbs, 314, 391–394 translations comparative wordlists, 66 machine (see machine translation) treebanks, 315–317 trees, 279–281 representing chunks, 270 traversal of, 280 trie, 162 trigram taggers, 204 truth conditions, 368 truth-conditional semantics in first

The Myth of Artificial Intelligence: Why Computers Can't Think the Way We Do

by Erik J. Larson  · 5 Apr 2021

labs at MIT, Stanford, and elsewhere were encountering seemingly endless quandaries, difficulties, confusions, and outright failures. The problem, for example, of Fully Automated High-Quality Machine Translation was thought in the 1950s to be solvable, given sufficient research effort and dollars. By the 1960s, government investment in translation dried up, on the

…

achieved general intelligence, and thus be capable of thinking and acting like humans. By the 1960s, therefore, the target of AI was the task of machine translation—the fully automatic rendering of texts from one language, such as Russian, into another, such as English. AI was “all in.” NATURAL LANGUAGE UNDERSTANDING As

…

Man can do.” Marvin Minsky, too, declared in 1967 that “within a generation, the problem of creating ‘artificial intelligence’ will be substantially solved.”2 But machine translation was a different ballgame, as researchers soon discovered. Having begun with a simplistic assumption, that language could be understood by analyzing words in large texts

…

high quality. Even programs working in specific domains such as biomedical literature were not fail-proof, and the failures were often embarrassingly incorrect and mindless. Machine translation researchers, in response, expanded their approach by exploring methods for “parsing” sentences, or finding syntactic structure in them, using new and powerful “transformational” grammars developed

…

had walked into a game of three-dimensional chess thinking it was tic-tac-toe.3 The National Resource Council (NRC) was pouring millions into machine-translation work at a number of American universities by the mid-1960s, but as for actual successes in engineering systems to understand, or even to simulate

…

the depth of problems the field faced.4 The effects of his reports on the research community were seismic. He had pinpointed the exact obstacle machine translation was foundering on, and it was irritatingly “philosophical”: the dearth of so-called common sense or “world knowledge”—knowledge about the actual world. Consider a

…

the NRC, the notion that computers could be programmed with the world knowledge of humans was “utterly chimerical, and hardly deserves any further discussion.”6 Machine translation was stuck, in other words, with results that were a far cry from fully automatic, high-quality translations (and that remain so today, although the

…

to data-driven or “empirical” methods seemed to liberate AI from those early, cloudy days of work on machine translation, when seemingly endless problems with capturing meaning and context plagued engineering efforts. In fact, machine translation itself was later cracked by a group of IBM researchers using a statistical (that is, not grammar-based

…

intensive—statistical approach pioneered by IBM Research Labs. SUCCESS … OR NOT The success of contemporary systems like Google Translate on the once puzzling problem of machine translation is often touted as evidence that AI will succeed, given enough time and the right ideas. The truth is more sobering. While it turns out

…

statistical or machine learning approaches, the original concerns of Bar-Hillel and others regarding semantics (meaning) and pragmatics (context) have proven to be well-founded. Machine translation, which had seemed like a difficult natural language problem, could be adequately accomplished with simple statistical analysis, given large corpora (datasets) in different languages. (And

…

note that machine translation is still not very high quality—it is more like “good enough to be useful.”) This is not evidence of impressive growth in machines’ natural

…

language understanding intelligence, but only evidence that machine translation is a much simpler problem than it was initially perceived to be. Again, deep problems with understanding language using computers have persisted. A simple way

…

MYCIN, which provided sometimes quite good medical diagnoses, made clear that AI methods were relevant to a variety of problems normally requiring high human intelligence. Machine translation, as we’ve seen, was an initial failure, but yielded to different approaches made possible by the availability of large datasets (a precursor to many

…

. Did the bot trick some judges some of the time? Yes. Did it actually pass the Turing test in any meaningful way? No.10 Though machine translation has in recent years been tamed (somewhat) by large volumes of texts translated into different languages on the web, the Turing test remains a perpetual

…

in the pen or I loved the river. I walked to the bank.) Confusingly, work in the 1960s on so-called fully-automated high-quality machine translation began with statistical approaches, albeit simpler ones. They didn’t work very well, and soon apostates like Yehoshua Bar-Hillel concluded that automatic translation was

…

IBM researchers, and web giants like Google used similar techniques (and now deep learning) to provide decent statistically-driven translation services. Fully-automated high-quality machine translation came full circle. But Bar-Hillel’s original skepticism is still germane; translations requiring knowledge and context are still a black box to modern approaches

…

Research Council, Publication 1416, 1966. 5. Sergei Nirenburg, H. L. Somers, and Yorick Wilks, eds., Readings in Machine Translation (Cambridge, MA: MIT Press, 2003), 75. 6. For a readable discussion of early problems with machine translation, see John Haugeland, Artificial Intelligence, The Very Idea (Cambridge, MA: MIT Press, 1989). Yehoshua Bar-Hillel’s comment

…

formal systems, 284n6 Frankenstein (fictional character), 238 Frankenstein: Or, a Modern Prometheus (novel, Shelly), 238, 280 frequency assumptions, 150–154, 173 Fully Automated High-Quality Machine Translation, 48 functions, 139 Galileo, 160 gambler’s fallacy, 122 games, 125–126 Gardner, Dan, 69–70 Garland, Alex, 79, 80, 289n16 Gates, Bill, 75 general

The Most Human Human: What Talking With Computers Teaches Us About What It Means to Be Alive

by Brian Christian  · 1 Mar 2011  · 370pp  · 94,968 words

of “writing a book”: this notion of style versus content, and of singularity and uniqueness of vision, is at the heart of recent debates about machine translation, especially of literature. Wolfram Alpha researcher and chatbot author Robert Lockhart describes the chatbot community as being split between two competing approaches, what he calls

…

happen along the way or prove to be an unnecessary middleman. This divide also plays out in the history of computer translation. For many decades, machine translation projects attempted to understand language in a rule-based way, breaking down a sentence’s structure and getting down to the underlying, universal meaning, before

…

re-encoding that meaning according to another language’s rules. In the 1990s, a statistical approach to machine translation—the approach that Google uses—came into its own, which left the question of meaning entirely out of it. Cleverbot, for instance, can know that

…

for (1) unusual and/or nonliteral phrasing, and (2) long-term consistency in point of view and style. On both of those counts, even as machine translation increasingly penetrates the world of business, literary novels remain mostly untranslatable by machine. What this also suggests, intriguingly, is that the task of translating (or

…

, rule-based paradigm—that is, up until about the turn of the century. If you isolate a specific type of problem—say, the problem of machine translation—you see the narrative clear as day. Early approaches were about building huge “dictionaries” of word-to-word pairings, based on meaning, and algorithms for

…

the translation problem. A new approach abandons these strategies, more or less entirely. For instance, the 2006 NIST machine translation competition was convincingly won by a team from Google, stunning a number of machine translation experts: not a single human on the Google team knew the languages (Arabic and Chinese) used in the competition

…

to do a pretty good simulation of translating other regurgitations of the same ilk … For works that are truly original—and therefore worth translating—statistical machine translation hasn’t got a hope.” 16. See, e.g., Columbia University clinical psychologist George Bonanno’s “Loss, Trauma, and Human Resilience: Have We Underestimated the

Thinking Machines: The Inside Story of Artificial Intelligence and Our Race to Build the Future

by Luke Dormehl  · 10 Aug 2016  · 252pp  · 74,167 words

the World’s Fair’s IBM exhibit and a powerful IBM mainframe computer 114 miles away in Kingston, New York, carrying out the heavy lifting. Machine translation was a simple, but brilliant, summation of how computers’ clear-thinking vision would usher us towards utopia. The politicians may not have been able to

…

AI designed to diagnose diseases more effectively than many human physicians. Finally, the miraculous World’s Fair Machine Translator is most familiar to us today as Google Translate: a free service which offers impressively accurate probabilistic machine translation between some fifty-eight different languages – or 3,306 separate translation services in total. If the

…

the next two decades, Larry Page and Sergey Brin’s company would make some of the biggest advances in AI history. These spanned fields including machine translation, pattern recognition, computer vision, autonomous robots and far more, which AI researchers had struggled with for half a century. Virtually none of it was achieved

…

playing a key role in discovering new types of medicine, making information accessible and useful to people around the world, allowing for quick and easy machine translation, and much, much more. Minsky might have prematurely dismissed neural networks before they truly took over, but other ideas of his remain in heavy circulation

…

as a microcosm for the larger question of how we recreate intelligence inside a machine. Today’s goal posts are hazier. Is Artificial Intelligence about machine translation, image recognition, teaching cars to autonomously drive themselves, smart assistants capable of running our lives, intelligent thermostats able to talk to your equally smart TV

We Are Data: Algorithms and the Making of Our Digital Selves

by John Cheney-Lippold  · 1 May 2017  · 420pp  · 100,811 words

, in this case the numbers one through five in English and Spanish. Source: Tomas Mikolov, Quoc Le, and Ilya Sutskever, “Exploiting Similarities among Languages for Machine Translation,” technical report, arXiv:1309.4168, 2. But the mathematical relationship between “manos” and “nigga” didn’t come from UN or EU documents. There is no

…

, Is That a Fish in Your Ear? Translation and the Meaning of Everything (New York: Farrar, Straus and Giroux, 2012). 70. Franz Joseph Och, “Statistical Machine Translation: Foundations and Recent Advances,” Google, September 12, 2005, www.mt-archive.info; Christian Boitet, Hervé Blanchon, Mark Seligman, and Valérie Bellynck, “Evolution of MT with

…

the Web,” Proceedings of the Conference “Machine Translation 25 Years On,” 2009, 1–13. 71. Franz Josef Och, “Selection and Use of Nonstatistical Translation Components in a Statistical

…

Machine Translation Framework,” Google, 2014, www.google.com. 72. Additionally, some languages also neocolonially move through their “closest” language to get to “English.” Catalan passes through Spanish,

AI in Museums: Reflections, Perspectives and Applications

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

Only Humans Need Apply: Winners and Losers in the Age of Smart Machines

by Thomas H. Davenport and Julia Kirby  · 23 May 2016  · 347pp  · 97,721 words

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

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

GCHQ

by Richard Aldrich  · 10 Jun 2010  · 826pp  · 231,966 words

Ghost Work: How to Stop Silicon Valley From Building a New Global Underclass

by Mary L. Gray and Siddharth Suri  · 6 May 2019  · 346pp  · 97,330 words

Machine: A White Space Novel

by Elizabeth Bear  · 5 Oct 2020  · 537pp  · 146,610 words

Children of Ruin

by Adrian Tchaikovsky  · 13 May 2019  · 471pp  · 147,210 words

New Dark Age: Technology and the End of the Future

by James Bridle  · 18 Jun 2018  · 301pp  · 85,263 words

Beautiful Data: The Stories Behind Elegant Data Solutions

by Toby Segaran and Jeff Hammerbacher  · 1 Jul 2009

Masterminds of Programming: Conversations With the Creators of Major Programming Languages

by Federico Biancuzzi and Shane Warden  · 21 Mar 2009  · 496pp  · 174,084 words

Architects of Intelligence

by Martin Ford  · 16 Nov 2018  · 586pp  · 186,548 words

Talk on the Wild Side

by Lane Greene  · 15 Dec 2018  · 284pp  · 84,169 words

The Information: A History, a Theory, a Flood

by James Gleick  · 1 Mar 2011  · 855pp  · 178,507 words

Structure and Interpretation of Computer Programs, Second Edition

by Harold Abelson, Gerald Jay Sussman and Julie Sussman  · 1 Jan 1984  · 1,387pp  · 202,295 words

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

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

I'm Feeling Lucky: The Confessions of Google Employee Number 59

by Douglas Edwards  · 11 Jul 2011  · 496pp  · 154,363 words

Hands-On Machine Learning With Scikit-Learn and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems

by Aurélien Géron  · 13 Mar 2017  · 1,331pp  · 163,200 words

The Globotics Upheaval: Globalisation, Robotics and the Future of Work

by Richard Baldwin  · 10 Jan 2019  · 301pp  · 89,076 words

The Cultural Logic of Computation

by David Golumbia  · 31 Mar 2009  · 268pp  · 109,447 words

The Creativity Code: How AI Is Learning to Write, Paint and Think

by Marcus Du Sautoy  · 7 Mar 2019  · 337pp  · 103,522 words

Future Politics: Living Together in a World Transformed by Tech

by Jamie Susskind  · 3 Sep 2018  · 533pp

Artificial Unintelligence: How Computers Misunderstand the World

by Meredith Broussard  · 19 Apr 2018  · 245pp  · 83,272 words

Machine Learning Design Patterns: Solutions to Common Challenges in Data Preparation, Model Building, and MLOps

by Valliappa Lakshmanan, Sara Robinson and Michael Munn  · 31 Oct 2020

Life After Google: The Fall of Big Data and the Rise of the Blockchain Economy

by George Gilder  · 16 Jul 2018  · 332pp  · 93,672 words

The Alignment Problem: Machine Learning and Human Values

by Brian Christian  · 5 Oct 2020  · 625pp  · 167,349 words

Genius Makers: The Mavericks Who Brought A. I. To Google, Facebook, and the World

by Cade Metz  · 15 Mar 2021  · 414pp  · 109,622 words

The Age of AI: And Our Human Future

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

Rule of the Robots: How Artificial Intelligence Will Transform Everything

by Martin Ford  · 13 Sep 2021  · 288pp  · 86,995 words

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

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

Literary Theory for Robots: How Computers Learned to Write

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

Who Owns This Sentence?: A History of Copyrights and Wrongs

by David Bellos and Alexandre Montagu  · 23 Jan 2024  · 305pp  · 101,093 words

Free Speech: Ten Principles for a Connected World

by Timothy Garton Ash  · 23 May 2016  · 743pp  · 201,651 words

The Theory That Would Not Die: How Bayes' Rule Cracked the Enigma Code, Hunted Down Russian Submarines, and Emerged Triumphant From Two Centuries of Controversy

by Sharon Bertsch McGrayne  · 16 May 2011  · 561pp  · 120,899 words

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

by Jordan Ellenberg  · 14 May 2021  · 665pp  · 159,350 words

Reinventing Discovery: The New Era of Networked Science

by Michael Nielsen  · 2 Oct 2011  · 400pp  · 94,847 words

Eternity

by Greg Bear  · 2 Jan 1988  · 523pp  · 129,580 words

Natural Language Annotation for Machine Learning

by James Pustejovsky and Amber Stubbs  · 14 Oct 2012  · 502pp  · 107,510 words

Fluent Forever: How to Learn Any Language Fast and Never Forget It

by Gabriel Wyner  · 4 Aug 2014  · 366pp  · 87,916 words

Final Jeopardy: Man vs. Machine and the Quest to Know Everything

by Stephen Baker  · 17 Feb 2011  · 238pp  · 77,730 words

Is That a Fish in Your Ear?: Translation and the Meaning of Everything

by David Bellos  · 10 Oct 2011  · 396pp  · 107,814 words

The Zero Marginal Cost Society: The Internet of Things, the Collaborative Commons, and the Eclipse of Capitalism

by Jeremy Rifkin  · 31 Mar 2014  · 565pp  · 151,129 words

Overcomplicated: Technology at the Limits of Comprehension

by Samuel Arbesman  · 18 Jul 2016  · 222pp  · 53,317 words

Human Compatible: Artificial Intelligence and the Problem of Control

by Stuart Russell  · 7 Oct 2019  · 416pp  · 112,268 words

Doing Data Science: Straight Talk From the Frontline

by Cathy O'Neil and Rachel Schutt  · 8 Oct 2013  · 523pp  · 112,185 words

In Our Own Image: Savior or Destroyer? The History and Future of Artificial Intelligence

by George Zarkadakis  · 7 Mar 2016  · 405pp  · 117,219 words

Facebook: The Inside Story

by Steven Levy  · 25 Feb 2020  · 706pp  · 202,591 words

Robot Rules: Regulating Artificial Intelligence

by Jacob Turner  · 29 Oct 2018  · 688pp  · 147,571 words

Talk to Me: How Voice Computing Will Transform the Way We Live, Work, and Think

by James Vlahos  · 1 Mar 2019  · 392pp  · 108,745 words

The Codebreakers: The Comprehensive History of Secret Communication From Ancient Times to the Internet

by David Kahn  · 1 Feb 1963  · 1,799pp  · 532,462 words

AIQ: How People and Machines Are Smarter Together

by Nick Polson and James Scott  · 14 May 2018  · 301pp  · 85,126 words

Text Analytics With Python: A Practical Real-World Approach to Gaining Actionable Insights From Your Data

by Dipanjan Sarkar  · 1 Dec 2016

I, Warbot: The Dawn of Artificially Intelligent Conflict

by Kenneth Payne  · 16 Jun 2021  · 339pp  · 92,785 words

The Singularity Is Nearer: When We Merge with AI

by Ray Kurzweil  · 25 Jun 2024

Kingdom of Characters: The Language Revolution That Made China Modern

by Jing Tsu  · 18 Jan 2022  · 408pp  · 105,715 words

The Last Lingua Franca: English Until the Return of Babel

by Nicholas Ostler  · 23 Nov 2010  · 484pp  · 120,507 words

Automate This: How Algorithms Came to Rule Our World

by Christopher Steiner  · 29 Aug 2012  · 317pp  · 84,400 words

The uplift war

by David Brin  · 1 Jun 1987  · 789pp  · 213,716 words

Deep Thinking: Where Machine Intelligence Ends and Human Creativity Begins

by Garry Kasparov  · 1 May 2017  · 331pp  · 104,366 words

Race Against the Machine: How the Digital Revolution Is Accelerating Innovation, Driving Productivity, and Irreversibly Transforming Employment and the Economy

by Erik Brynjolfsson  · 23 Jan 2012  · 72pp  · 21,361 words

Deep Medicine: How Artificial Intelligence Can Make Healthcare Human Again

by Eric Topol  · 1 Jan 2019  · 424pp  · 114,905 words

Paper Knowledge: Toward a Media History of Documents

by Lisa Gitelman  · 26 Mar 2014

WTF?: What's the Future and Why It's Up to Us

by Tim O'Reilly  · 9 Oct 2017  · 561pp  · 157,589 words

Track Changes

by Matthew G. Kirschenbaum  · 1 May 2016  · 519pp  · 142,646 words

The Big Nine: How the Tech Titans and Their Thinking Machines Could Warp Humanity

by Amy Webb  · 5 Mar 2019  · 340pp  · 97,723 words

Designing Data-Intensive Applications: The Big Ideas Behind Reliable, Scalable, and Maintainable Systems

by Martin Kleppmann  · 16 Mar 2017  · 1,237pp  · 227,370 words

Superminds: The Surprising Power of People and Computers Thinking Together

by Thomas W. Malone  · 14 May 2018  · 344pp  · 104,077 words

The Inevitable: Understanding the 12 Technological Forces That Will Shape Our Future

by Kevin Kelly  · 6 Jun 2016  · 371pp  · 108,317 words

Found in Translation: How Language Shapes Our Lives and Transforms the World

by Nataly Kelly and Jost Zetzsche  · 1 Oct 2012  · 274pp  · 73,344 words

Designing Data-Intensive Applications: The Big Ideas Behind Reliable, Scalable, and Maintainable Systems

by Martin Kleppmann  · 17 Apr 2017

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

by Parmy Olson  · 284pp  · 96,087 words

The Invisible Web: Uncovering Information Sources Search Engines Can't See

by Gary Price, Chris Sherman and Danny Sullivan  · 2 Jan 2003  · 481pp  · 121,669 words

What to Think About Machines That Think: Today's Leading Thinkers on the Age of Machine Intelligence

by John Brockman  · 5 Oct 2015  · 481pp  · 125,946 words

Nerds on Wall Street: Math, Machines and Wired Markets

by David J. Leinweber  · 31 Dec 2008  · 402pp  · 110,972 words

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

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

The AI Economy: Work, Wealth and Welfare in the Robot Age

by Roger Bootle  · 4 Sep 2019  · 374pp  · 111,284 words

The Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence

by Kate Crawford  · 5 Apr 2021  · 355pp  · 89,587 words

The Wealth of Humans: Work, Power, and Status in the Twenty-First Century

by Ryan Avent  · 20 Sep 2016  · 323pp  · 90,868 words

Ten Billion Tomorrows: How Science Fiction Technology Became Reality and Shapes the Future

by Brian Clegg  · 8 Dec 2015  · 315pp  · 92,151 words

Who Owns the Future?

by Jaron Lanier  · 6 May 2013  · 510pp  · 120,048 words

Big Data: A Revolution That Will Transform How We Live, Work, and Think

by Viktor Mayer-Schonberger and Kenneth Cukier  · 5 Mar 2013  · 304pp  · 82,395 words

Our Final Invention: Artificial Intelligence and the End of the Human Era

by James Barrat  · 30 Sep 2013  · 294pp  · 81,292 words

The Caryatids

by Bruce Sterling  · 24 Feb 2009  · 387pp  · 105,250 words

A World Without Work: Technology, Automation, and How We Should Respond

by Daniel Susskind  · 14 Jan 2020  · 419pp  · 109,241 words

The Seventh Sense: Power, Fortune, and Survival in the Age of Networks

by Joshua Cooper Ramo  · 16 May 2016  · 326pp  · 103,170 words

The Industries of the Future

by Alec Ross  · 2 Feb 2016  · 364pp  · 99,897 words

Code Dependent: Living in the Shadow of AI

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

Why Machines Learn: The Elegant Math Behind Modern AI

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

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

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

A New History of the Future in 100 Objects: A Fiction

by Adrian Hon  · 5 Oct 2020  · 340pp  · 101,675 words

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

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

Superintelligence: Paths, Dangers, Strategies

by Nick Bostrom  · 3 Jun 2014  · 574pp  · 164,509 words

More Money Than God: Hedge Funds and the Making of a New Elite

by Sebastian Mallaby  · 9 Jun 2010  · 584pp  · 187,436 words

Structure and interpretation of computer programs

by Harold Abelson, Gerald Jay Sussman and Julie Sussman  · 25 Jul 1996  · 893pp  · 199,542 words

Mastering Structured Data on the Semantic Web: From HTML5 Microdata to Linked Open Data

by Leslie Sikos  · 10 Jul 2015

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

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

Stealth

by Peter Westwick  · 22 Nov 2019  · 474pp  · 87,687 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

Narrative Economics: How Stories Go Viral and Drive Major Economic Events

by Robert J. Shiller  · 14 Oct 2019  · 611pp  · 130,419 words

The Man Who Invented the Computer

by Jane Smiley  · 18 Oct 2010  · 253pp  · 80,074 words

The Future of the Professions: How Technology Will Transform the Work of Human Experts

by Richard Susskind and Daniel Susskind  · 24 Aug 2015  · 742pp  · 137,937 words

Accelerando

by Stross, Charles  · 22 Jan 2005  · 489pp  · 148,885 words

Future Crimes: Everything Is Connected, Everyone Is Vulnerable and What We Can Do About It

by Marc Goodman  · 24 Feb 2015  · 677pp  · 206,548 words

AI Superpowers: China, Silicon Valley, and the New World Order

by Kai-Fu Lee  · 14 Sep 2018  · 307pp  · 88,180 words

Sparks: China's Underground Historians and Their Battle for the Future

by Ian Johnson  · 26 Sep 2023  · 407pp  · 119,073 words

There's a War Going on but No One Can See It

by Huib Modderkolk  · 1 Sep 2021  · 295pp  · 84,843 words

The AGI Chronicles

by Kevin Roose  · 481pp  · 151,124 words

Cage of Souls

by Adrian Tchaikovsky  · 4 Apr 2019  · 703pp  · 196,052 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

On Language: Chomsky's Classic Works Language and Responsibility and Reflections on Language in One Volume

by Noam Chomsky and Mitsou Ronat  · 26 Jul 2011

Pattern Recognition

by William Gibson  · 2 Jan 2003  · 385pp  · 99,985 words

The Formula: How Algorithms Solve All Our Problems-And Create More

by Luke Dormehl  · 4 Nov 2014  · 268pp  · 75,850 words

Because Internet: Understanding the New Rules of Language

by Gretchen McCulloch  · 22 Jul 2019  · 413pp  · 106,479 words

Calling Bullshit: The Art of Scepticism in a Data-Driven World

by Jevin D. West and Carl T. Bergstrom  · 3 Aug 2020

Babel No More: The Search for the World's Most Extraordinary Language Learners

by Michael Erard  · 10 Jan 2012  · 392pp  · 104,760 words

Data-Ism: The Revolution Transforming Decision Making, Consumer Behavior, and Almost Everything Else

by Steve Lohr  · 10 Mar 2015  · 239pp  · 70,206 words

Noam Chomsky: A Life of Dissent

by Robert F. Barsky  · 2 Feb 1997

Possible Minds: Twenty-Five Ways of Looking at AI

by John Brockman  · 19 Feb 2019  · 339pp  · 94,769 words

The Dream Machine: J.C.R. Licklider and the Revolution That Made Computing Personal

by M. Mitchell Waldrop  · 14 Apr 2001

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

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

The Mysterious Mr. Nakamoto: A Fifteen-Year Quest to Unmask the Secret Genius Behind Crypto

by Benjamin Wallace  · 18 Mar 2025  · 431pp  · 116,274 words

Human + Machine: Reimagining Work in the Age of AI

by Paul R. Daugherty and H. James Wilson  · 15 Jan 2018  · 523pp  · 61,179 words

Devil's Bargain: Steve Bannon, Donald Trump, and the Storming of the Presidency

by Joshua Green  · 17 Jul 2017  · 296pp  · 78,112 words

Information: A Very Short Introduction

by Luciano Floridi  · 25 Feb 2010  · 137pp  · 36,231 words

The Fractalist

by Benoit Mandelbrot  · 30 Oct 2012

The Mind Is Flat: The Illusion of Mental Depth and the Improvised Mind

by Nick Chater  · 28 Mar 2018  · 263pp  · 81,527 words

Red Moon

by Kim Stanley Robinson  · 22 Oct 2018  · 492pp  · 141,544 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 We Owe the Future: A Million-Year View

by William MacAskill  · 31 Aug 2022  · 451pp  · 125,201 words

Simple Rules: How to Thrive in a Complex World

by Donald Sull and Kathleen M. Eisenhardt  · 20 Apr 2015  · 294pp  · 82,438 words

The Economic Singularity: Artificial Intelligence and the Death of Capitalism

by Calum Chace  · 17 Jul 2016  · 477pp  · 75,408 words

Applied Artificial Intelligence: A Handbook for Business Leaders

by Mariya Yao, Adelyn Zhou and Marlene Jia  · 1 Jun 2018  · 161pp  · 39,526 words

The Internet Trap: How the Digital Economy Builds Monopolies and Undermines Democracy

by Matthew Hindman  · 24 Sep 2018

The Business of Platforms: Strategy in the Age of Digital Competition, Innovation, and Power

by Michael A. Cusumano, Annabelle Gawer and David B. Yoffie  · 6 May 2019  · 328pp  · 84,682 words

The Reverse Centaur's Guide to Life After AI: How to Think About Artificial Intelligence—Before It's Too Late

by Cory Doctorow  · 22 Jun 2026  · 175pp  · 51,295 words

This Will Make You Smarter: 150 New Scientific Concepts to Improve Your Thinking

by John Brockman  · 14 Feb 2012  · 416pp  · 106,582 words

You Are What You Speak: Grammar Grouches, Language Laws, and the Politics of Identity

by Robert Lane Greene  · 8 Mar 2011  · 319pp  · 95,854 words

Surviving AI: The Promise and Peril of Artificial Intelligence

by Calum Chace  · 28 Jul 2015  · 144pp  · 43,356 words

Smarter Than Us: The Rise of Machine Intelligence

by Stuart Armstrong  · 1 Feb 2014  · 48pp  · 12,437 words

Blue Ocean Strategy, Expanded Edition: How to Create Uncontested Market Space and Make the Competition Irrelevant

by W. Chan Kim and Renée A. Mauborgne  · 20 Jan 2014  · 287pp  · 80,180 words

Picnic Comma Lightning: In Search of a New Reality

by Laurence Scott  · 11 Jul 2018  · 244pp  · 81,334 words

Collaborative Society

by Dariusz Jemielniak and Aleksandra Przegalinska  · 18 Feb 2020  · 187pp  · 50,083 words

Covid-19: The Pandemic That Never Should Have Happened and How to Stop the Next One

by Debora MacKenzie  · 13 Jul 2020  · 266pp  · 80,273 words

The Future Is Asian

by Parag Khanna  · 5 Feb 2019  · 496pp  · 131,938 words

How Doctors Think

by Jerome Groopman  · 15 Jan 2007  · 292pp  · 94,324 words

Why Things Bite Back: Technology and the Revenge of Unintended Consequences

by Edward Tenner  · 1 Sep 1997

The Left Case Against the EU

by Costas Lapavitsas  · 17 Dec 2018  · 221pp  · 46,396 words

Spike: The Virus vs The People - The Inside Story

by Jeremy Farrar and Anjana Ahuja  · 15 Jan 2021  · 245pp  · 71,886 words

Language and Mind

by Noam Chomsky  · 1 Jan 1968

Understanding Sponsored Search: Core Elements of Keyword Advertising

by Jim Jansen  · 25 Jul 2011  · 298pp  · 43,745 words

Pandora's Brain

by Calum Chace  · 4 Feb 2014  · 345pp  · 104,404 words