spb/wp44_uqo

1% ============================================================================2% Part II — The Evolution of Computer Science3% Written by Agent L (Academic Writer), WP44, 2026-11-03.4% Bibliography: references_part02.bib (every key verified; see references_verification_part02.csv).5% Corpus quantities only through the macros of tables/numbers.tex; \pending{} marks missing macros.6% ============================================================================78\section{The disciplinary structure of computer science}9\label{sec:p2-structure}1011Any claim that computer science has been transformed since late 2022 presupposes a map of the discipline on which the transformation can be located. That map is not given by nature. Computer science is organised simultaneously by its objects (processors, programs, data, networks, users), by its methods (proof, construction, measurement, learning), and by the communities and venues through which results circulate; the three organisations overlap only partially, and every bibliometric classification is a compromise among them. This section describes the compromise adopted in the present study and the alternatives against which it must be read. It serves two purposes for the remainder of the manuscript: it fixes the vocabulary of \emph{areas} and \emph{categories} used in Parts~V--IX, and it makes explicit the sense in which a ``shift of research toward artificial intelligence'' is a measurable proposition rather than a slogan.1213\subsection{Classification schemes and what they encode}14\label{sec:p2-schemes}1516Four classification systems are relevant to a corpus of the kind assembled here, and they encode the structure of the discipline in different ways.1718The \emph{ACM Computing Classification System} (CCS) is the only one produced by a professional society as an explicit ontology. Its 2012 revision replaced the 1998 four-level scheme with a poly-hierarchy of roughly two thousand concepts organised under thirteen top-level nodes, among them ``Theory of computation'', ``Computing methodologies'' (which houses artificial intelligence, machine learning, computer graphics and simulation side by side), ``Computer systems organization'', ``Software and its engineering'', ``Security and privacy'', ``Human-centered computing'' and ``Networks'' \citep{rous2012ccs}. The CCS is a statement of how the ACM understood the field in 2012. It is applied by authors at submission time, is used almost exclusively in ACM venues, and is therefore sparse in any corpus that spans IEEE, Springer, Elsevier, open-access publishers and preprint servers. Its value for this study is conceptual rather than operational: it provides an independent, expert-designed reference against which the forty categories of CLAUDE.md~\S4 can be positioned, and it documents a pre-transformer view in which large language models, foundation models, AI agents and AI safety do not exist as concepts at all.1920The \emph{arXiv subject classes} are the second system. arXiv was created in 1991 for high-energy physics and its taxonomy reflects that origin; the computer science archive (\texttt{cs}) is organised into forty subject classes largely inherited from the ACM scheme of the 1990s \citep{ginsparg2011arxiv}. Authors choose a primary class and may add cross-lists, so the classification is self-reported, multi-label, and coarse. Its distinctive property is temporal: a preprint is classified at the moment of submission, before peer review, so the arXiv class records how authors positioned their work at the time, not how a later indexer reconstructed it. For the questions of this paper that property matters: when a paper on, say, program verification with language models is submitted to \texttt{cs.PL} with a cross-list to \texttt{cs.LG}, the pair of classes is a small but genuine piece of evidence about disciplinary boundaries as practitioners perceive them.2122The \emph{DBLP computer science bibliography} is the third. DBLP does not classify papers at all; it classifies \emph{venues}. Begun in 1993 as a database and logic programming bibliography and extended to the whole of computer science, it records the tables of contents of conferences and journals with curated author identities \citep{ley2002dblp}. Membership in DBLP is itself a judgement that a venue belongs to computer science, and the venue stream to which a paper belongs is a strong signal of community: a paper in the proceedings of OSDI is an operating-systems paper in the sociological sense whatever its subject matter. DBLP therefore encodes the discipline as a set of communities rather than as a set of topics, and the two encodings disagree precisely where this paper expects change, namely where machine-learning methods enter venues that were not built for them.2324The fourth system is the family of \emph{algorithmic, citation-based classifications} that underlie modern open bibliographic indexes. OpenAlex, which succeeded the Microsoft Academic Graph after the latter was discontinued at the end of 2021 \citep{wang2020mag,priem2022openalex}, assigns to every work a ranked list of \emph{topics} from a fixed vocabulary of about 4{,}500 topics grouped into subfields, fields and domains; the topics were obtained by clustering the citation network and labelling the clusters with a language model, and are assigned to new works by a classifier over title, abstract and venue \citep{openalex2025topics}. The approach descends from the publication-level clustering methodology of \citet{waltman2012classification}, refined by the Leiden community-detection algorithm \citep{traag2019leiden}, and it inherits the known properties of that family: clusters are well connected and reproducible, but their granularity is a free parameter with no natural value \citep{sjogarde2018granularity}, and citation-based taxonomies, while more accurate than text-based or journal-based ones on average \citep{klavans2017taxonomy}, lag behind the literature by construction because a cluster must accumulate citations before it can exist. The Semantic Scholar academic graph offers a comparable but differently constructed resource \citep{kinney2023s2}. Older, journal-based classifications of the kind proposed by \citet{glanzel2003classification} remain in use in evaluative scientometrics but are too coarse, and too journal-centric, for a conference-dominated discipline.2526Each of these systems is a model of the discipline with its own inductive bias. The CCS encodes expert judgement frozen in 2012; arXiv classes encode author self-positioning at submission; DBLP streams encode venue communities; OpenAlex topics encode citation neighbourhoods as of a recent snapshot. None of them is ground truth, and this study treats their disagreements as data rather than as noise.2728\subsection{The forty-category ontology of this study}29\label{sec:p2-ontology}3031The classification used throughout the manuscript is the hierarchical ontology specified in CLAUDE.md~\S4 and built in Phase~2 (Appendix~A). It comprises forty leaf categories grouped into eight areas: artificial intelligence and machine learning (nine leaves, from large language models to AI safety); perception, language, robotics and interaction (six); software, programming languages and formal methods (four); theory of computation (two); data, information and distributed systems (six); security, privacy and trust (five); hardware, architecture and emerging computing (five); and computational science and discovery (three). A small number of \emph{specialisation} edges record that one leaf refines another: large language models and multimodal AI refine foundation models; compilers refine programming languages; cloud and edge computing refine distributed computing; federated learning refines privacy-preserving computation; AI accelerators refine computer architecture. The areas are a reporting device; classification operates on leaves.3233The assignment rule is deliberately simple so that it can be audited. Let $\mathcal{W}$ be the set of works and $\mathcal{C}$ the set of forty leaves. For each leaf $c$ the ontology lists weighted signals of five types $k \in \{\text{topic},\ \text{subfield},\ \text{arXiv},\ \text{stream},\ \text{keyword}\}$, with signal set $S_k(c)$ and weights $\omega_{kj} \in (0,1]$ for $j \in S_k(c)$. A work $w$ matches a signal $j$ if the corresponding attribute of $w$ (an OpenAlex topic identifier, a subfield, an arXiv class, a DBLP stream, or a regular expression over title and abstract) contains it. The score of $w$ for $c$ is34\begin{equation}35s(w,c) \;=\; \sum_{k} \min\Bigl\{1,\ \sum_{j \in S_k(c)} \omega_{kj}\, \mathbf{1}[w \text{ matches } j]\Bigr\} \;-\; \infty \cdot \mathbf{1}[w \text{ matches an exclusion of } c],36\label{eq:p2-score}37\end{equation}38so that no single signal type can contribute more than one unit, a weight of one means ``this signal alone suffices'', and an exclusion pattern vetoes the leaf outright. The work belongs to every leaf with $s(w,c) \geq 1$; its \emph{primary} leaf is $\arg\max_c s(w,c)$, with ties broken in favour of a child over its parent and then by category number. Multi-label membership is not an accident of the rule but its purpose: an instruction-tuning paper is expected to score in both large language models and natural language processing, and the share of works with more than one leaf is itself one of the quantities tracked in Part~IX.3940Two design choices deserve comment because they shape every downstream result. First, the ontology is not a relabelling of OpenAlex topics. Six AI categories that the citation-based topics cannot separate (large language models, foundation models, multimodal AI, generative AI, AI agents, AI safety) and seven small categories with no dedicated topic or venue (federated learning, sustainable computing, autonomous scientific discovery, privacy-preserving computation, blockchain, quantum computing, edge computing) rely on keyword patterns of weight one; the build fails if such a pattern is absent. The reason is historical and is the subject of Section~\ref{sec:p2-waves}: the citation clusters that OpenAlex labels were formed by a literature in which ``large language model'' was not yet a stable term, and a topic such as ``Topic Modeling'' in fact hosts most BERT-era and retrieval-augmented work. Second, the ontology is positioned relative to the CCS rather than derived from it. The AI area corresponds to the CCS branch \emph{Computing methodologies $\to$ Artificial intelligence / Machine learning}, but the leaves inside it are drawn from the post-2017 literature; theory, systems, networks, security and human-centred computing map onto the corresponding CCS top-level nodes almost one-to-one; computational biology, scientific computing and sustainable computing sit in CCS under \emph{Applied computing} and \emph{Social and professional topics}, which the present ontology promotes to first-class areas because the questions of this paper concern precisely the traffic between computing and the sciences it serves.4142The ontology's limitations are stated in Appendix~A and are not repeated here, except for one that bears on interpretation. A category defined partly by keywords is defined partly by vocabulary, and vocabulary changes faster than research practice. If the phrase ``large language model'' spreads into papers whose methods are unchanged, a keyword-driven category will register growth that is partly nominal. Section~\ref{sec:p2-claims} returns to this problem under the heading of LLM-assisted writing; Part~IV describes the embedding-based checks designed to detect it.4344\subsection{Where the classifications disagree}45\label{sec:p2-disagree}4647The disagreements among the four systems are systematic, and they fall into three kinds.4849The first is \emph{coverage}. Comparisons of bibliographic databases consistently show that no source indexes the whole of a discipline and that the gaps are not random: conference proceedings, preprints, and non-English venues are the most unevenly covered \citep{visser2021comparison,martinmartin2021coverage}. OpenAlex, the backbone of this corpus, inherits the broad but uneven coverage of Microsoft Academic and Crossref; its reference lists are less complete than those of the commercial indexes for part of the period, which affects citation-based measures more than counts \citep{culbert2025openalex}. The linking phase of this study illustrates the scale of the problem at the level of record identity. The OpenAlex snapshot contains \LinkOaCsRecords{} works classified in computer science between 2018 and 2026; DBLP contributes \LinkDblpTwoZeroOneEightTwoZeroTwoSix{} records over the same years, of which \LinkDblpLinkedDistinct{} could be matched to an OpenAlex work by DOI or by title and year in the first linking pass and \LinkDblpUnlinked{} could not; the arXiv \texttt{cs} harvest contributes \LinkArxivCsRecords{} preprints, of which \LinkArxivLinkedDistinct{} were matched, overwhelmingly through the arXiv identifier that OpenAlex stores. These are linkage statistics, not findings about research, but they establish that a classification applied to one source cannot simply be transferred to another: a non-trivial fraction of the venue-defined (DBLP) universe has no citation-defined (OpenAlex) counterpart, and vice versa. Part~III describes the second linking pass and the treatment of unmatched records.5051The second kind of disagreement is \emph{granularity and boundary}. The arXiv class \texttt{cs.LG} is one node among forty; in OpenAlex, machine learning spreads across dozens of topics in several fields, including statistics, engineering and medicine, because the citation network does not respect the boundary of the computer-science field. Conversely, OpenAlex's field~17 (``Computer Science'') contains clusters on educational technology, media studies and physics that no computer scientist would claim, and the ontology of this study deliberately leaves most of them unmapped (Appendix~A). The practical consequence is that the apparent size of ``AI'' depends on which boundary is used: a venue-based definition (DBLP streams of NeurIPS, ICML, ICLR, AAAI and the vision and language conferences) gives a different set from a topic-based definition, which gives a different set again from an arXiv-class definition, and the three sets diverge most in the applied sciences, where machine-learning methods are used without the paper belonging to any machine-learning community. The multi-label design of the ontology, and the use of all three signal types with capped weights in equation~\eqref{eq:p2-score}, is the response to this divergence: a work that is an AI paper by every criterion scores high on all of them, while a work that is an AI paper by vocabulary alone is held to a higher evidentiary standard.5253The third kind is \emph{temporal}. The CCS was fixed in 2012 and the arXiv classes are older still; OpenAlex topics were generated from a citation network observed in the early 2020s and are reassigned when the snapshot is rebuilt; DBLP streams are stable but new venues appear and old ones are renamed. A classification that does not change cannot register the emergence of a field except as the growth of an existing node; a classification that is regenerated cannot be compared across snapshots without care. Neither property is a defect, but they pull in opposite directions, and a study of structural change must choose. This study freezes all classifications at a single snapshot (Part~III), so that any growth it reports is growth of a fixed category, and it treats the emergence of \emph{new} categories as a separate detection problem (Part~IV, Section~\ref{sec:p2-claims}).5455A final, sociological disagreement should be named explicitly because it affects what counts as a ``paper''. Computer science is unusual among the sciences in treating selective conferences as the primary venue of record, a practice that has been defended on grounds of speed and criticised on grounds of reviewing quality and the fragmentation of the literature \citep{vardi2009conferences,fortnow2009growup}. The convention has two consequences for bibliometrics. Conference papers are indexed less uniformly than journal articles, so coverage disagreements are concentrated in exactly the venues that the discipline values most; and the conference calendar imposes a strong seasonality on publication dates, so that monthly series must be deseasonalised before any break can be read off them. The preprint culture adds a third consequence. Computer scientists came to arXiv late relative to physicists and mathematicians, and a large share of computer-science preprints are never matched to a published version, either because they are not published or because the match fails \citep{lin2020preprints}. A corpus that counts preprints and published versions as separate works will over-count the preprint-heavy subfields; a corpus that links them will under-count whenever linkage fails; and the subfields where linkage fails most often are, again, the fast-moving ones.5657\subsection{Why the growth of a field is hard to measure}58\label{sec:p2-growth}5960The questions of this paper are ultimately about rates: of publication, of citation, of movement between categories. Rates presuppose denominators, and the denominators of science are themselves contested quantities.6162The literature on the growth of science offers two sobering lessons. The first is that measured growth depends heavily on the database. Using cited references as a proxy for the literature that scientists actually use, \citet{bornmann2015growth} estimated the long-run growth of science at roughly three per cent per year with a doubling time of about a quarter of a century, while estimates based on the publication counts of individual databases vary widely with the database's coverage policy and its own expansion. A later analysis with segmented growth curves across several databases found that the growth rate has itself changed over time, with faster growth in recent decades, but that the location of the segments depends on the source \citep{bornmann2021growth}. For the present study the lesson is that a change in the volume of computer-science publications between 2018 and 2026 is the sum of a change in research activity and a change in indexing, and that only the first is of scientific interest. OpenAlex grew substantially during the period as it absorbed new sources and improved its matching; the indexing lag (the delay between a work's publication and its appearance in the snapshot) is longest for the most recent months, so that any series ending in 2026 slopes downward at its end for reasons that have nothing to do with research. These are among the pre-period diagnostics and sensitivity checks required by CLAUDE.md~\S6.5 and specified in Part~IV.6364The second lesson concerns the ``science of science'' as a field in its own right. The programme described by \citet{fortunato2018sos} and developed at book length by \citet{wang2021sos} has established robust regularities: citation distributions are heavy-tailed within every field; team size has grown steadily; novelty in the sense of atypical combinations of prior work is associated with higher impact at the top of the distribution; and the productivity of individual scientists is approximately stable over careers. Those regularities are the background against which any field-specific claim must be read. A rise in the citation inequality of AI papers after 2022, for instance, is interesting only if it exceeds the inequality that heavy-tailed citation distributions produce in every growing field; the hypotheses H1--H8 of the preregistered protocol (Part~III) are formulated as comparisons between the post-2022 period and the baseline for that reason.6566A further difficulty is specific to computer science and concerns what the discipline takes itself to be. \citet{denning2005science} argued that computer science is a science of information processes and not merely an engineering of artefacts, and \citet{wing2006computational} that ``computational thinking'' is a general intellectual skill rather than a professional one; both arguments were written at a moment when the discipline was defending its boundaries. The post-2022 period raises the converse question: whether the discipline's methods have diffused so widely that the boundary is no longer where the classifications place it. Papers that apply learned models to protein structure, materials, weather or law are classified as computer science by some systems and not others; a transformation that consists partly of such diffusion is partly invisible to a classification that keeps the boundary fixed. The ontology's eighth area, computational science and discovery, and the ``penumbra'' set defined in Part~III (works with a computer-science topic but without venue or arXiv support) exist to make that diffusion observable without inflating the core.6768\section{Three prior transformations and how they propagated}69\label{sec:p2-waves}7071The post-2022 period is not the first time computer science has been reorganised around a new technology. Three earlier transformations are directly relevant because they established the publication practices, the infrastructure, and the conceptual vocabulary within which the present one unfolds: the internet and web era, which created the systems and the industrial research laboratories; the deep-learning wave of 2012--2017, which created the benchmark culture and the compute economy; and the transformer and pretraining era of 2017--2022, which created the models themselves. For each, the account below is deliberately narrow. It does not attempt a history of computing; it asks what changed in how research was done and published, and how the change propagated from the subfield in which it began to the others. The answers suggest a way of reading the post-2022 data that the empirical parts of this paper then test.7273\subsection{The internet and web era}74\label{sec:p2-web}7576The World-Wide Web was presented to the computing community in 1994 as an information system built on three modest conventions, a naming scheme, a transfer protocol and a markup language, that happened to compose \citep{bernerslee1994www}. Within five years the problem of finding things on the web had produced two results that still structure information retrieval and network science: the use of the hyperlink graph as a source of authority, through PageRank in the Google prototype \citep{brin1998anatomy} and through the hubs-and-authorities formulation of \citet{kleinberg1999hubs}. Within another five, the engineering problem of running such systems at scale had produced the Google File System and MapReduce \citep{ghemawat2003gfs,dean2008mapreduce}, and within a decade the commoditisation of that infrastructure as ``cloud computing'' had been analysed as a shift in the economics of computing rather than in its technology \citep{armbrust2010cloud}. Apache Spark, which generalised MapReduce into a unified engine for iterative and interactive workloads, marks the maturation of the era \citep{zaharia2016spark}.7778Three features of this transformation matter for what followed. First, the decisive results came from industrial laboratories or from academic groups that immediately founded companies, and they were published in the discipline's selective conferences (SOSP, OSDI, WWW, SIGCOMM) rather than in journals; the practice by which an industrial systems paper describes a deployed artefact, with measurements but without a released implementation, dates from this period and persists in the large-model reports discussed below. Second, the web made open-source software the ordinary form of a systems research artefact: Hadoop, the open re-implementation of MapReduce, propagated the method more effectively than the paper did, and the later machine-learning frameworks follow the same pattern. Third, the era created the data. Web-scale text, images, logs and graphs were the precondition for statistical methods in language, vision and recommendation, and the subfields that had access to such data (information retrieval, speech, machine translation) were the first to adopt statistical learning, a decade before the deep-learning wave made it general.7980The propagation pattern was \emph{infrastructural}. The web changed the objects of study in networks, databases, distributed systems and security, and it created new subfields (web search, social computing, cloud computing), but it did not change the methods of theory, programming languages, graphics or human-computer interaction except by supplying them with new problems. The discipline's map was extended rather than redrawn: the ACM classification of 2012 could still accommodate every web-era subfield under its existing top-level nodes \citep{rous2012ccs}.8182\subsection{The deep-learning wave, 2012--2017}83\label{sec:p2-dl}8485The second transformation is usually dated to 2012, when a convolutional network trained on two GPUs reduced the top-5 error on the ImageNet Large Scale Visual Recognition Challenge by a margin that no incremental improvement in hand-designed features could match \citep{krizhevsky2017imagenet,deng2009imagenet,russakovsky2015imagenet}, and when deep networks replaced Gaussian mixtures in the acoustic models of the major speech laboratories at roughly the same time \citep{hinton2012speech}. The methods were not new: the long short-term memory cell dates from 1997 \citep{hochreiter1997lstm}, convolutional networks from the 1980s, and the retrospective of \citet{lecun2015deep} presents the decade's results as the delayed payoff of a thirty-year programme. What was new was the conjunction of labelled data at the scale of ImageNet, general-purpose graphics processors, and a benchmark with a public leaderboard.8687The subsequent five years produced the architectural vocabulary that the transformer era would inherit. Distributed word representations trained by simple predictive objectives \citep{mikolov2013word2vec} made it natural to treat language as a sequence of dense vectors; the variational autoencoder \citep{kingma2014vae} and the generative adversarial network \citep{goodfellow2014gan} established deep generative modelling as a research programme; sequence-to-sequence learning with recurrent encoders and decoders \citep{sutskever2014seq2seq} and, crucially, the attention mechanism that let a decoder consult every encoder state \citep{bahdanau2015attention} reframed machine translation and then every conditional generation task; residual connections \citep{he2016resnet} removed the practical ceiling on depth. In reinforcement learning, deep Q-networks learned Atari games from pixels \citep{mnih2015dqn}, and the AlphaGo lineage combined deep networks with tree search to defeat professional Go players, first with human games as training data and then without them \citep{silver2016alphago,silver2017alphagozero}, generalising to chess and shogi with a single algorithm \citep{silver2018alphazero} and, in a different formulation, to real-time strategy games \citep{vinyals2019alphastar}.8889The changes in publication practice were as consequential as the methods. Four deserve to be stated because the post-2022 data will be read against them.9091The first is the \emph{benchmark as the unit of progress}. ImageNet established the pattern in which a public dataset, a fixed metric and a leaderboard define a subfield's agenda; the pattern was reproduced in language understanding, question answering and reinforcement learning, and it made progress legible to outsiders at the cost of concentrating effort on whatever the benchmark measured. The critiques of that culture, which argue that benchmarks are chosen for convenience rather than validity and that leaderboard competition rewards narrow optimisation \citep{dehghani2021lottery,raji2021benchmark}, were written in the transformer era but describe a practice born in this one.9293The second is the \emph{preprint-first, conference-second} publication cycle. Machine-learning results began to appear on arXiv months before, or instead of, their conference versions; the practice spread from a minority to a norm within the decade, and it is the reason the arXiv \texttt{cs} harvest is indispensable to this corpus. The accompanying costs were catalogued at the end of the period: speculation presented as explanation, mathematics as decoration, suggestive language and the conflation of explanation with description \citep{lipton2019troubling}. Concerns about reproducibility were raised at the same time, both in audits of what published papers actually report \citep{gundersen2018reproducibility} and in the organised reproducibility programmes that the major conferences introduced in response \citep{pineau2021reproducibility}.9495The third is the \emph{open-source framework as a shared instrument}. TensorFlow and PyTorch \citep{abadi2016tensorflow,paszke2019pytorch}, released by Google and Facebook respectively, lowered the cost of entry to deep learning to the point where a method published on arXiv could be reproduced and extended within days; they also standardised the computational substrate on NVIDIA GPUs and, later, on tensor processing units \citep{jouppi2017tpu}, which made compute a measurable input to research. The retrospective analysis of \citet{sevilla2022compute} identifies this period as the start of a ``deep learning era'' in which the training compute of notable systems grew with a doubling time of about six months, far faster than before 2010; \citet{thompson2020limits} argued that such growth is economically and environmentally unsustainable, a theme to which the sustainable-computing category of this study returns.9697The fourth is the \emph{entry of industry as a producer rather than a consumer of research}. The laboratories of Google, DeepMind, Facebook, Microsoft and Baidu published the decisive results of the period in the academic venues and on arXiv, and their access to data and compute began to separate them from universities. \citet{ahmed2020dedemocratization} documented the growing presence of large technology firms and elite universities at the major AI conferences and the relative decline of other institutions; the later and broader analysis of \citet{ahmed2023industry} found that industry produced an increasing share of the largest models, the most-cited papers and the leading benchmarks. Whether that concentration intensified after 2022 is hypothesis H6 of the protocol.9899The propagation pattern of this wave was \emph{methodological} and sequential. Deep learning entered computer vision and speech first, natural language processing second (roughly 2014--2016, with the sequence-to-sequence results), reinforcement learning and robotics third, and then, more slowly and with more resistance, the subfields whose objects are not perceptual: program analysis \citep{allamanis2018survey}, systems (learned index structures and learned scheduling), databases, and scientific computing \citep{raissi2019pinn}. In each case the receiving subfield first treated learning as a tool for a pre-existing problem, and only later, if at all, reorganised its problems around it. That two-stage pattern, tool adoption followed by problem redefinition, is the template against which Part~IX reads the post-2022 movement of methods across the forty categories.100101\subsection{The transformer and pretraining era, 2017--2022}102\label{sec:p2-transformer}103104The third transformation began with an architecture and ended with a change in what a model is for. The Transformer of \citet{vaswani2017attention} replaced recurrence with self-attention in the encoder--decoder framework of neural machine translation; its immediate merit was parallelism in training, and its lasting one was that the same block could be stacked to any depth and applied to any sequence of tokens. Within eighteen months it had been combined with the idea, already present in ELMo's contextual word vectors \citep{peters2018elmo}, that a language model trained on unlabelled text could be \emph{pretrained} once and \emph{fine-tuned} on many tasks: BERT \citep{devlin2019bert}, with its masked-language objective, set new results on every task of the GLUE benchmark \citep{wang2018glue} and on SQuAD \citep{rajpurkar2016squad}, and its successors showed that most of its design choices were less important than the amount of data and training \citep{liu2019roberta,raffel2020t5}. The generative branch of the family, the GPT series of decoder-only models, reached its third member in 2020 with 175 billion parameters and the demonstration that a sufficiently large language model could perform new tasks from a few examples placed in its prompt, without any gradient update \citep{brown2020gpt3}. In-context learning, more than any architectural novelty, is what separates the transformer era from the deep-learning wave: the unit of research shifted from the task-specific model to the general model and the prompt.105106The era had a quantitative theory of its own growth. \citet{kaplan2020scaling} reported that the test loss of autoregressive language models falls as a power law in parameters, data and compute over several orders of magnitude, with the implication that larger models are more sample-efficient; \citet{hoffmann2022chinchilla} corrected the allocation, showing that models of the GPT-3 generation were undertrained relative to their size and that compute-optimal training requires data and parameters to scale together. Between the two papers, laboratories built models of increasing size on the strength of the first law (Gopher, PaLM, OPT and the open multilingual BLOOM \citep{rae2021gopher,chowdhery2022palm,zhang2022opt,bigscience2022bloom}), and the systems research needed to train them, model parallelism, memory-partitioned optimisers and IO-aware attention kernels \citep{shoeybi2019megatron,rajbhandari2020zero,dao2022flashattention}, became a research area in its own right, as did sparse mixtures of experts that decouple parameter count from per-token compute \citep{fedus2022switch}. The claim that certain abilities appear discontinuously with scale \citep{wei2022emergent} became one of the era's most cited and most contested propositions; the companion observation that prompting a model to write out intermediate reasoning improves its answers on arithmetic and symbolic tasks \citep{wei2022cot} established prompting as a method. New benchmarks were built at the scale of the models: a fifty-seven-subject examination of knowledge \citep{hendrycks2021mmlu}, a harder successor to GLUE \citep{wang2019superglue}, and a collaborative suite of more than two hundred tasks explicitly designed to resist saturation \citep{srivastava2023bigbench}.107108The transformer did not remain a language architecture. Vision transformers showed that image patches could be treated as tokens and that, with enough data, the inductive bias of convolution was dispensable \citep{dosovitskiy2021vit}; self-supervised objectives, contrastive \citep{chen2020simclr}, self-distilled \citep{caron2021dino} and masked \citep{he2022mae}, removed the dependence on labels. Contrastive language--image pretraining aligned the two modalities in a shared embedding space and produced zero-shot classifiers from natural-language prompts \citep{radford2021clip}; in the same months, autoregressive text-to-image generation \citep{ramesh2021dalle} and then diffusion models, a decade-old idea \citep{sohl2015diffusion} made practical by denoising objectives \citep{ho2020ddpm}, score-based formulations \citep{song2021score} and architectural refinements that overtook adversarial models \citep{dhariwal2021beats}, converged on text-conditioned image synthesis at photographic quality \citep{ramesh2022dalle2,saharia2022imagen}. The latent-diffusion formulation \citep{rombach2022ldm}, released with open weights in the summer of 2022 under the name Stable Diffusion, is the generative-AI counterpart of ChatGPT: a system whose public availability, more than its technical novelty, changed who could use it. Speech followed the self-supervised path \citep{baevski2020wav2vec}; few-shot visual language models \citep{alayrac2022flamingo} and a single network trained across text, images, games and robot control \citep{reed2022gato} prefigured the multimodal and agentic systems of the following period. In structural biology, a transformer-based system solved, for most single-chain proteins, a problem that had resisted fifty years of effort \citep{jumper2021alphafold}, and was quickly joined by an independent three-track network \citep{baek2021rosettafold}.109110Two further lines of work prepared the specific object that appeared in November 2022. Models trained on public code could synthesise functions from docstrings \citep{chen2021codex}, which turned software engineering into an early adopter of large models and created the first commercial product built on one. And the problem of making a language model do what a user asks, rather than what the pretraining distribution predicts, was addressed by reinforcement learning from human preferences, a method developed for control tasks \citep{christiano2017rlhf}, applied to summarisation \citep{stiennon2020summarize}, and then used to fine-tune GPT-3 into InstructGPT \citep{ouyang2022instructgpt}, whose 1.3-billion-parameter version was preferred by raters to the 175-billion-parameter base model. The same method, framed around helpfulness and harmlessness, was being developed in parallel elsewhere \citep{bai2022hh}. Retrieval-augmented generation, which conditions a generator on documents retrieved from an external index, was proposed in the same period \citep{lewis2020rag} and would become the standard answer to the models' factual unreliability.111112The era also produced its own critique. The report that coined the term ``foundation model'' argued that the homogenisation of research around a few pretrained models was a change in the structure of the field, with risks that were social and institutional as much as technical \citep{bommasani2021foundation}; the ``stochastic parrots'' paper questioned whether ever-larger language models were a sound research direction at all, on environmental, documentary and epistemic grounds \citep{bender2021parrots}. The environmental argument had been quantified for NLP models by \citet{strubell2019energy}, contested and refined for the largest models by \citet{patterson2021carbon}, and turned into a research agenda by \citet{schwartz2020green}; the carbon footprint of a 176-billion-parameter model was measured end to end by \citet{luccioni2023bloom}.113114What changed in publication practice during this era can be summarised in three observations. The leading models were described in \emph{technical reports} rather than in peer-reviewed papers, often with hundreds of authors, sometimes without the training data, the code, or the weights; the preprint became the version of record for the most consequential work in the discipline. The infrastructure for sharing models became an institution in its own right: the Transformers library \citep{wolf2020transformers} and the model hub built around it made the pretrained model, not the paper, the thing that propagated. And the compute required to participate at the frontier grew beyond the reach of universities, so that the concentration documented by \citet{ahmed2023industry} is, in this era, a concentration of capability rather than merely of attention.115116\subsection{A comparative reading of the three waves}117\label{sec:p2-compare}118119Table~\ref{tab:p2-waves} summarises the three transformations along the dimensions that matter for the empirical design of this study. The comparison is qualitative; the corpus-level counterpart is the subject of Part~IX.120121\begin{table}[htbp]122\centering123\caption{Three prior transformations of computer science compared on the dimensions relevant to the post-2022 analysis. Qualitative characterisation by the authors from the works cited in Sections~\ref{sec:p2-web}--\ref{sec:p2-transformer}; no corpus quantities.}124\label{tab:p2-waves}125\small126\begin{tabularx}{\textwidth}{@{}lXXX@{}}127\toprule128 & Internet and web (c.~1994--2010) & Deep learning (2012--2017) & Transformer and pretraining (2017--2022) \\129\midrule130Originating subfields & Networks, information retrieval, distributed systems & Computer vision, speech & Natural language processing \\131Core artefact & Protocols, search engines, warehouse-scale systems & Trained task-specific networks; frameworks & Pretrained general models; prompts; weights \\132Locus of production & Industry and start-ups; academic systems groups & Academic groups with industry laboratories entering & Industry laboratories, with academic participation declining at the frontier \\133Publication practice & Selective conferences; deployed-artefact papers; open-source re-implementations & arXiv-first; leaderboards; shared frameworks; reproducibility programmes & Technical reports; model hubs; very large author lists; preprint as version of record \\134Propagation pattern & Infrastructural: new objects and subfields, methods of other areas unchanged & Methodological and sequential: tool adoption then problem redefinition, perceptual fields first & Substitutive: one architecture and one pretraining recipe across modalities; general models replace task-specific ones \\135Visibility in classifications & Accommodated by existing top-level nodes & New venues and arXiv classes gain weight; topics split & Vocabulary outruns citation clusters; keyword-defined categories required \\136\bottomrule137\end{tabularx}138\end{table}139140Three conclusions follow from the comparison and shape the hypotheses tested later. First, the mechanisms of propagation differ in kind. The web changed the problems of other subfields; deep learning changed their methods; the transformer era changed the \emph{models} they use, which can be adopted without changing either problems or methods. A subfield can therefore register the third transformation at low cost, by using a pretrained model as a component, and the resulting growth in the vocabulary of the subfield need not indicate a reorganisation of its research. Distinguishing component adoption from reorganisation is the purpose of the citation-flow and embedding analyses of Parts~IV and~VIII.141142Second, each wave shortened the publication cycle and shifted the version of record toward the preprint. The consequence for this study is that the arXiv harvest and the OpenAlex index measure different things: the former records when a result was first made public, the latter when it entered the citation system. A break detected in one series and not in the other is informative about the mechanism, not merely about measurement.143144Third, the three waves were cumulative in infrastructure and in institutions. The industrial laboratories that produced the web systems trained the generation that produced deep learning, and the frameworks and compute of the deep-learning wave made the transformer era possible. The post-2022 period inherited all of it: the laboratories, the frameworks, the model hubs, the benchmark culture, and the preprint norm. Whatever is new about the period must be identified against that inheritance rather than attributed to a single product release, and the baseline window of 2018--2022 was chosen so that the inheritance is observed directly rather than assumed.145146\section{The non-AI areas at the baseline, 2018--2022}147\label{sec:p2-baseline}148149The hypotheses of this study concern not only the growth of AI but the condition of everything else. H2 predicts that at least one non-AI area declines in relative share while growing in absolute terms; H3 predicts that formal verification, computational complexity, operating systems and computer architecture show \emph{no} structural break in the window around November 2022; H4 predicts that citations crossing from those areas into AI increase. Such predictions can be evaluated only if the state of each area at the baseline is described independently of the AI narrative. This section provides that description for the thirty-one non-AI leaves of the ontology, grouped by area, in a deliberately compressed form: for each, the problems that defined the 2018--2022 literature, the landmark results that the corpus's citation data should recover, and the points of contact with machine learning that existed before 2022. Where a result is known to the authors only from a report rather than from a peer-reviewed paper, that is stated.150151\subsection{Systems, cloud and distributed computing}152\label{sec:p2-systems}153154The systems literature of the baseline period was organised around the consequences of the warehouse-scale computer \citep{barroso2018datacenter}. Three agendas dominated. The first was the \emph{serverless} abstraction, in which the unit of deployment is a stateless function and the provider manages everything else; the Berkeley view presented it as the next step in the simplification of cloud programming \citep{jonas2019serverless}, while a contemporaneous critique from the database community argued that its restrictions on state, communication and hardware access made it a step backward for data-intensive work \citep{hellerstein2019serverless}. The disagreement was productive: much of the 2019--2022 systems literature consists of attempts to give serverless platforms the state and the locality they lacked. The second agenda was \emph{resource disaggregation}, the separation of compute, memory and storage into independently scaled pools connected by fast networks, an idea with roots in blade-server memory expansion \citep{lim2009disaggregated} that became a central topic of operating-systems research once network latency fell within an order of magnitude of memory latency. The third was \emph{tail latency}: in a service composed of thousands of components, rare slow responses in any of them dominate the user-visible distribution, and the techniques of hedged requests, replication and tied requests described by \citet{dean2013tail} became the standard vocabulary of large-scale service design.155156Operating systems research in the period was marked by the consequences of the Spectre and Meltdown disclosures, discussed under security below, which forced kernel designers to reconsider the cost of isolation, and by the growth of in-kernel programmability. Distributed computing as a theory remained focused on consensus, fault tolerance and the coordination of replicated state, with blockchain systems supplying new adversarial models (Section~\ref{sec:p2-security}). The contact with machine learning before 2022 was of two kinds: ``systems for ML'', the training and serving infrastructure of Section~\ref{sec:p2-transformer}, which systems venues absorbed readily, and ``ML for systems'', the use of learned models for scheduling, configuration and resource management, which was an active but minority topic. The ontology treats cloud and edge computing as children of distributed computing precisely because the venue structure does: OSDI, SOSP, EuroSys and NSDI publish all three.157158\subsection{Hardware and computer architecture}159\label{sec:p2-hardware}160161The baseline period in computer architecture was dominated by the end of two scaling regimes. Dennard scaling, the observation that transistor power density stays constant as transistors shrink \citep{dennard1974scaling}, had ended in the mid-2000s; its consequence, that a growing fraction of a chip must be left dark at any time, was quantified by \citet{esmaeilzadeh2011dark} and set the agenda for the following decade. The slowing of Moore's law completed the picture. The Turing lecture of \citet{hennessy2019golden} drew the conclusion that general-purpose performance gains were over and that the future lay in domain-specific architectures, open instruction sets and agile hardware design; \citet{leiserson2020room} argued in parallel that performance would henceforth come from software, algorithms and hardware specialisation rather than from the ``bottom'' of the stack. The tensor processing unit \citep{jouppi2017tpu} is the canonical example of the domain-specific architecture, and its publication in a mainstream architecture venue is a measure of how completely machine learning had become the architecture community's defining workload by 2017. AI accelerators are a child of computer architecture in the ontology for that reason, and the hypothesis that the architecture category shows no post-2022 break is, in effect, the hypothesis that this absorption was complete before ChatGPT.162163Compilers underwent a parallel reorganisation around machine-learning workloads. The multi-level intermediate representation MLIR \citep{lattner2021mlir}, designed to let one compiler infrastructure serve domain-specific languages, tensor programs and hardware targets alike, and the end-to-end tensor compiler TVM \citep{chen2018tvm} represent the state of the art at the baseline. Both treat the deep-learning graph as a first-class compilation target; neither depends on large language models.164165\subsection{Theory of computation}166\label{sec:p2-theory}167168Two developments characterise theoretical computer science at the baseline, one internal and one at the boundary with machine learning.169170The internal development is \emph{fine-grained complexity}: the programme of explaining why polynomial-time problems such as edit distance, all-pairs shortest paths and orthogonal vectors have resisted improvement for decades, by proving that faster algorithms for one would contradict a small number of hardness conjectures, above all the strong exponential-time hypothesis \citep{vassilevska2015hardness,williams2018finegrained}. The programme reorganised algorithms research around conditional lower bounds in the same way that NP-completeness had organised it around unconditional intractability, and its vocabulary is a reliable marker of the complexity category in this corpus.171172The boundary development is the \emph{theory of deep learning}. The empirical observation that networks with enough parameters to memorise random labels nevertheless generalise \citep{zhang2021rethinking} undermined the classical uniform-convergence account of generalisation and prompted new theory: the double-descent curve, which reconciles the classical bias--variance trade-off with the behaviour of overparameterised models \citep{belkin2019reconciling}, and the neural tangent kernel, which characterises the training dynamics of infinitely wide networks \citep{jacot2018ntk}. The machine-learning-theory leaf of the ontology houses this literature, which belongs sociologically to the learning conferences and intellectually to the theory community. Its presence is a reminder that the theory area was in contact with AI long before 2022, through learning theory, optimisation and the complexity of sampling and inference.173174\subsection{Programming languages, compilers and formal verification}175\label{sec:p2-pl}176177The programming-languages literature of the baseline is marked by the success of a language. Rust, whose ownership type system gives memory safety without garbage collection \citep{matsakis2014rust}, became the first systems language to be adopted at scale on the strength of a safety guarantee; the RustBelt project supplied the semantic foundation by proving, in a separation-logic framework mechanised in Coq, that the safe fragment is sound and that the unsafe library code on which it rests respects the invariants \citep{jung2017rustbelt}. The result is characteristic of the period's formal-methods culture: a verification effort that follows an engineering artefact rather than preceding it, and that uses a proof assistant as its medium.178179Proof assistants themselves reached a new stage. The verified C compiler CompCert \citep{leroy2009compcert} and the verified microkernel seL4 \citep{klein2009sel4} had shown a decade earlier that production software could be proved correct; the Kepler conjecture was formally proved in 2017 after a twelve-year effort \citep{hales2017kepler}; and the Lean theorem prover \citep{demoura2015lean}, re-implemented as a programming language in its fourth version \citep{demoura2021lean4}, acquired a community-maintained mathematical library whose scale and organisation were new in formalised mathematics \citep{mathlib2020}. The relevance to the post-2022 period is direct: Lean and its library became the environment in which language models were later asked to prove theorems, so that formal verification, one of the categories predicted \emph{not} to break in H3, is also one where the contact with AI is most visible in the recent literature. The hypothesis is therefore a sharp one.180181Software engineering had already begun to absorb learned models before 2022. The survey of \citet{allamanis2018survey} organised a decade of ``big code'' research around the naturalness hypothesis, that source code is as statistically regular as natural language; the CodeXGLUE benchmark collection \citep{lu2021codexglue} standardised evaluation; and the Codex models \citep{chen2021codex} made program synthesis from natural language a product. The software-engineering category is thus expected to show an early and strong AI signal, and its baseline is not an AI-free one.182183\subsection{Security and cryptography}184\label{sec:p2-security}185186Three events organise the security and cryptography literature of the baseline. The first is the disclosure, in January 2018, of Spectre and Meltdown, which showed that speculative execution in essentially all high-performance processors leaks data across isolation boundaries \citep{kocher2019spectre,lipp2020meltdown}. The disclosures created a sustained literature on microarchitectural side channels and their mitigations that cuts across security, architecture and operating systems, and they illustrate the kind of cross-area citation flow that H4 measures.187188The second is the standardisation of \emph{post-quantum cryptography}. Shor's algorithm \citep{shor1997algorithms} had made the eventual obsolescence of RSA and elliptic-curve schemes a matter of engineering timelines, and the case for migrating before large quantum computers exist was made to a general scientific audience by \citet{bernstein2017pqc}. The process run by the United States National Institute of Standards and Technology from 2016 selected lattice-based and hash-based schemes, published as federal standards in August 2024 \citep{nist2024fips203,nist2024fips204,nist2024fips205}. The applied-cryptography category of this corpus therefore contains a large and growing body of work whose dynamics are driven by a standards calendar entirely independent of language models, which makes it a useful placebo field. Fully homomorphic encryption, shown to be possible by \citet{gentry2009fhe}, remained the subject of a steady literature on efficiency.189190The third is the maturation of \emph{adversarial machine learning}. The observation that imperceptible perturbations can change a network's output \citep{szegedy2014intriguing} and the attacks and evaluation standards that followed \citep{carlini2017robustness} created, by 2018, a field with ten years of history and a recognised research agenda \citep{biggio2018wild}. The ontology scores adversarial work in both cybersecurity and AI safety, which reflects the baseline state: the problem belongs to both communities and neither owns it. Blockchain systems, finally, entered the period with a systematised research agenda \citep{bonneau2015sok} and left it as a mature if contested area whose publication dynamics follow cryptocurrency markets more than academic calendars.191192\subsection{Privacy-preserving computation and federated learning}193\label{sec:p2-privacy}194195Differential privacy, defined by \citet{dwork2006calibrating} and systematised in the monograph of \citet{dwork2014algorithmic}, was by 2018 the accepted formal standard for statistical disclosure control, and its combination with stochastic gradient descent \citep{abadi2016dpsgd} had made private training of deep networks practical at modest cost in accuracy. Federated learning, proposed as a way to train models on data that never leaves the devices that generated it \citep{mcmahan2017fl}, grew in five years into a field with its own open problems in optimisation, systems, privacy and incentives, catalogued in a survey of unusual breadth \citep{kairouz2021fl}. Both categories are keyword-driven in the ontology because neither has a venue of its own; both were already AI-adjacent at the baseline, so that any post-2022 growth they show cannot be read as a new contact with AI.196197\subsection{Data systems, information retrieval and networks}198\label{sec:p2-data}199200Database research at the baseline was absorbing the proposal that learned models could replace the data structures at the core of the system: the learned index \citep{kraska2018learned} argued that a model of the cumulative distribution of keys can outperform a B-tree, and the argument was extended to query optimisation, cardinality estimation and tuning. In parallel, approximate nearest-neighbour search over dense vectors, long a niche concern, acquired the data structures and the GPU implementations that would later make vector databases an industry: hierarchical navigable small-world graphs \citep{malkov2020hnsw} and billion-scale similarity search on GPUs \citep{johnson2021faiss}. Information retrieval made the corresponding move from lexical to dense retrieval, in which queries and passages are embedded by a pretrained encoder and matched by inner product \citep{karpukhin2020dpr}; the retrieval-augmented generation of \citet{lewis2020rag} is the junction of that line with language modelling. Of all the non-AI areas, information retrieval is the one whose baseline most closely anticipates the post-2022 period, and its category should be read accordingly.201202Computer networks entered the period with programmability as its organising idea. OpenFlow had separated the control plane from the forwarding plane \citep{mckeown2008openflow}; the P4 language made the forwarding plane itself programmable \citep{bosshart2014p4}; and the QUIC transport, designed and deployed at internet scale by a single company before standardisation \citep{langley2017quic}, exemplified a mode of networking research in which deployment precedes publication. Edge computing, defined as the placement of computation near the sources of data \citep{shi2016edge}, grew with the internet of things and with the latency requirements of on-device inference; the ontology treats it as a child of distributed computing and shares its IoT topics with networks.203204\subsection{Human--computer interaction and computer graphics}205\label{sec:p2-hci}206207Human--computer interaction had, by 2019, a research programme on interaction with AI systems that predates the conversational interfaces of 2022. Guidelines for human--AI interaction were derived from two decades of literature and validated against deployed products \citep{amershi2019guidelines}; the particular difficulty of designing for systems whose capabilities are uncertain and whose outputs are open-ended was analysed by \citet{yang2020reexamining}; and the programme of human-centred AI, which places reliability, safety and trust above autonomy, was articulated by \citet{shneiderman2020hcai}. The HCI category is therefore one where a post-2022 surge, if observed, would be the intensification of an existing agenda rather than the arrival of a new one. Its venues (CHI, UIST, CSCW) are also among the most clearly delimited in DBLP, which makes the category's venue signal strong.208209Computer graphics underwent a transformation of its own in 2020, when neural radiance fields showed that a multilayer perceptron could represent a scene well enough to synthesise photorealistic novel views from a handful of photographs \citep{mildenhall2020nerf}. Neural rendering had been surveyed as an emerging area in the same year \citep{tewari2020neural}; within three years, explicit Gaussian primitives optimised by gradient descent had matched the quality of radiance fields at real-time frame rates \citep{kerbl2023gaussian}. The sequence is instructive for the detection methods of Part~IV: neural radiance fields are one of the pre-registered ``known emergence'' topics against which the emerging-topic detector is validated (H8), and they emerged in a non-AI category by the adoption of an AI method, which is exactly the pattern the detector must distinguish from vocabulary drift.210211\subsection{Quantum computing}212\label{sec:p2-quantum}213214Quantum computing at the baseline is defined by a term coined to describe it. \citet{preskill2018nisq} named the ``noisy intermediate-scale quantum'' era, in which devices of fifty to a few hundred qubits without error correction might perform useful tasks, and warned that no such task had yet been demonstrated. The claim of quantum computational advantage on a sampling problem \citep{arute2019supremacy} and the review of algorithms designed for such devices \citep{bharti2022nisq} bracket the period; the demonstration that logical error rates fall as a surface code is scaled \citep{google2023surface} and, later, that they fall below threshold \citep{google2025threshold} mark the transition toward fault tolerance that the NISQ framing anticipated. The quantum-computing category is small, keyword-driven, and vetoes physics papers that lack a computing term (Appendix~A); its dynamics are driven by hardware milestones and have no obvious mechanism of contact with language models, which makes it, with post-quantum cryptography, a natural placebo.215216\subsection{Scientific computing, computational biology and sustainable computing}217\label{sec:p2-science}218219Scientific computing entered the period through the door of physics-informed learning. Physics-informed neural networks, which impose a partial differential equation as a loss term \citep{raissi2019pinn}, and neural operators, which learn mappings between function spaces rather than between vectors \citep{li2021fno}, created a literature that the review of \citet{karniadakis2021piml} already presented as a field. Computational biology had its own transformation in the baseline window and does not need the post-2022 one to explain its trajectory: the first AlphaFold had won the 2018 structure-prediction assessment \citep{senior2020alphafold}, the second effectively settled the single-chain problem in 2020--2021 \citep{jumper2021alphafold}, protein language models predicted structure from sequence alone at scale \citep{lin2023esm}, and the third AlphaFold extended prediction to complexes with nucleic acids and ligands \citep{abramson2024af3}. The review of \citet{wang2023aiscience} in 2023 took stock of ``AI for science'' across disciplines; the autonomous-scientific-discovery leaf of the ontology, which requires an AI or automation term, is the place where the corpus records the next step, in which models participate in the design of experiments rather than in the analysis of their results.220221Sustainable computing, finally, was an established concern before 2022 but a small category. The careful recalibration of global data-centre energy use \citep{masanet2020datacenter} found that efficiency gains had largely offset growth in demand through 2018, a conclusion that the training and serving of large models would later put under pressure; the energy and carbon studies cited in Section~\ref{sec:p2-transformer} \citep{strubell2019energy,schwartz2020green,patterson2021carbon,luccioni2023bloom} connected the concern directly to AI. The category is keyword-driven and is expected to grow; whether its growth is driven by AI workloads or by the broader energy agenda is a question for Part~IX.222223Taken together, the baseline descriptions support a conclusion that the empirical parts must not forget: there was no AI-free area in 2022. Every non-AI category had a point of contact with machine learning before ChatGPT, ranging from marginal (complexity theory, post-quantum cryptography) to pervasive (information retrieval, software engineering, computational biology). The post-2022 question is therefore not whether AI \emph{entered} these areas but whether the rate and the kind of contact changed, and the baseline window exists so that the rate can be estimated before the reference date.224225\section{November 2022 as a reference date}226\label{sec:p2-nov2022}227228\subsection{What ChatGPT was and was not}229\label{sec:p2-chatgpt}230231On 30 November 2022 OpenAI released ChatGPT as a ``research preview'': a conversational interface to a language model of the GPT-3.5 series, fine-tuned by the reinforcement-learning-from-human-feedback method of InstructGPT, free to use through a web browser \citep{openai2022chatgpt}. No technical report accompanied the release; the announcement described the training procedure by reference to \citet{ouyang2022instructgpt} and listed limitations (plausible but incorrect answers, sensitivity to phrasing, verbosity, incomplete refusal of harmful requests) that the subsequent literature would study at length. It is important for the design of this paper to be precise about what was new. The underlying capabilities, few-shot prompting, instruction following, code generation, had been published between 2020 and 2022 \citep{brown2020gpt3,chen2021codex,ouyang2022instructgpt} and were available through a commercial programming interface. The architecture was five years old. What ChatGPT added was an interface that required no programming, a price of zero, and a conversational format that made the model's abilities and failures legible to anyone; its diffusion among the public was by all accounts faster than that of any previous software product, though the figures usually cited for it come from industry analysts rather than from verifiable sources and are not repeated here.232233For a study of the \emph{research} literature, this distinction has a consequence that is easily missed. The research community did not learn about large language models on 30 November 2022; it had been building and publishing them for three years. What changed on that date, if anything, was the attention of everyone else: funders, universities, firms, governments, and the researchers in subfields that had not previously had a reason to engage. If the date marks anything in the corpus, it should mark a change in \emph{who} publishes on language models and \emph{where}, more than a change in what the language-model community itself does. That expectation is formalised in H4 and H6 and in the reading of the cross-area citation flows in Part~VIII.234235\subsection{The releases before and after}236\label{sec:p2-timeline}237238ChatGPT was one release in a dense sequence, and the sequence matters because any of its members is a candidate reference date. Table~\ref{tab:p2-timeline} lists the ones whose primary documents could be verified. Within the twelve months before November 2022 came the compute-optimal scaling result \citep{hoffmann2022chinchilla}, the PaLM, OPT and BLOOM models \citep{chowdhery2022palm,zhang2022opt,bigscience2022bloom}, the chain-of-thought and emergent-abilities papers \citep{wei2022cot,wei2022emergent}, the second DALL-E and Imagen \citep{ramesh2022dalle2,saharia2022imagen}, the open release of Stable Diffusion \citep{rombach2022ldm}, the Flamingo and Gato systems \citep{alayrac2022flamingo,reed2022gato}, the Whisper speech model \citep{radford2023whisper}, and the ReAct formulation of language models that interleave reasoning with actions \citep{yao2023react}, posted in October 2022 and the direct ancestor of the agent literature. Within the twelve months after came LLaMA, the first competitive open-weight model family \citep{touvron2023llama}, Toolformer \citep{schick2023toolformer}, the GPT-4 technical report \citep{openai2023gpt4}, an early and much-debated qualitative evaluation of GPT-4 \citep{bubeck2023sparks}, Llama~2 with a permissive licence \citep{touvron2023llama2}, the RT-2 vision-language-action model that transferred web-scale pretraining to robot control \citep{brohan2023rt2}, Gaussian splatting in graphics \citep{kerbl2023gaussian}, the Mamba state-space architecture as the first serious challenger to attention \citep{gu2023mamba}, and the Gemini models \citep{gemini2023}. Later milestones, Llama~3 \citep{grattafiori2024llama3}, AlphaFold~3 \citep{abramson2024af3}, the o1 reasoning models \citep{openai2024o1} and DeepSeek-R1, which reported that reasoning behaviour could be elicited by reinforcement learning with verifiable rewards and released its weights \citep{deepseek2025r1}, fall inside the observation window and are candidate dates for the \emph{later} breaks that the changepoint methods of Part~IV may find.239240\begin{table}[htbp]241\centering242\caption{Candidate reference dates around November 2022. ``Document'' is the verified primary source; dates are those of the first public version of the document or of the release it describes, as given in the source. The list is restricted to items cited in this Part and is not a complete chronology.}243\label{tab:p2-timeline}244\small245\begin{tabularx}{\textwidth}{@{}llX@{}}246\toprule247Date & Release or result & Document \\248\midrule2492022-03 & Compute-optimal scaling (Chinchilla) & \citet{hoffmann2022chinchilla}, preprint \\2502022-04 & PaLM; DALL-E 2; Flamingo & \citet{chowdhery2022palm,ramesh2022dalle2,alayrac2022flamingo}, preprints \\2512022-05 & OPT; Gato; Imagen & \citet{zhang2022opt,reed2022gato,saharia2022imagen}, preprints \\2522022-06 & Latent diffusion published (CVPR); Stable Diffusion released openly in August & \citet{rombach2022ldm} \\2532022-10 & ReAct & \citet{yao2023react}, preprint \\2542022-11 & BLOOM; ChatGPT (30 November) & \citet{bigscience2022bloom}, preprint; \citet{openai2022chatgpt}, web page \\2552022-12 & Whisper & \citet{radford2023whisper}, preprint \\2562023-02 & LLaMA; Toolformer & \citet{touvron2023llama,schick2023toolformer}, preprints \\2572023-03 & GPT-4 & \citet{openai2023gpt4}, technical report (preprint) \\2582023-07 & Llama 2; RT-2 & \citet{touvron2023llama2,brohan2023rt2}, preprints \\2592023-08 & 3D Gaussian splatting (SIGGRAPH) & \citet{kerbl2023gaussian} \\2602023-12 & Mamba; Gemini & \citet{gu2023mamba,gemini2023}, preprints \\2612024-05 & AlphaFold 3 & \citet{abramson2024af3} \\2622024-07 & Llama 3 & \citet{grattafiori2024llama3}, preprint \\2632024-12 & o1 system card & \citet{openai2024o1}, preprint \\2642025-01 & DeepSeek-R1 & \citet{deepseek2025r1}, preprint \\265\bottomrule266\end{tabularx}267\end{table}268269Two features of the table are relevant to identification. The first is that the great majority of the documents are preprints or technical reports; none of the model releases of 2022--2025 was described in a peer-reviewed venue at the time of release, and several never were. A corpus that excluded preprints would be blind to the events it is trying to date. The second is that the events are not sparse. Between March 2022 and March 2023 there was, on average, a landmark release every month, and a time-series method that is free to choose the break date will find a break somewhere in that interval for reasons that have nothing to do with ChatGPT in particular. That is the subject of the next subsection.270271\subsection{Why the date is used descriptively and not causally}272\label{sec:p2-causal}273274The mandate of this study (CLAUDE.md~\S6.5) states that the November 2022 boundary is a historical reference point and not proof of a causal effect of ChatGPT. The statement is a constraint on interpretation, but it is also a methodological position that can be made precise.275276Let $N_{c,t}$ denote the number of works assigned to category $c$ in month $t$, and consider the simplest model that the ``post-ChatGPT'' framing suggests, an interrupted time series of the form used in programme evaluation \citep{bernal2017its}:277\begin{equation}278\log N_{c,t} \;=\; \alpha_c + \beta_c\, t + \delta_c\, \mathbf{1}[t \geq \tau] + \gamma_c\,(t-\tau)\,\mathbf{1}[t \geq \tau] + \mathbf{s}_c(t) + \varepsilon_{c,t},279\label{eq:p2-its}280\end{equation}281with $\tau$ fixed at December 2022, $\mathbf{s}_c(t)$ a seasonal component tied to the conference calendar, $\delta_c$ a level shift and $\gamma_c$ a change in trend. Three conditions must hold for $\delta_c$ or $\gamma_c$ to be read as the effect of the event at $\tau$: the pre-period trend must be well specified, so that an acceleration already under way is not attributed to the break; nothing else that affects $N_{c,t}$ may change at $\tau$; and the date $\tau$ must have been chosen independently of the data. The first condition is testable and is tested in Part~IX with pre-period trend diagnostics. The second fails on its face: Table~\ref{tab:p2-timeline} lists a dozen co-occurring events, the compute trends of \citet{sevilla2022compute} were already exponential, and the growth of the arXiv itself, documented month by month on its statistics page \citep{arxiv2026stats}, is a confounder that applies to every category. The third condition is the subtle one. The date was chosen because of what happened afterwards, which is the definition of selecting a break on the outcome; a test of a known break at $\tau$ \citep{bernal2017its} has the wrong size when $\tau$ was in fact selected from a set of candidates, and the appropriate tools are those for a break at an \emph{unknown} date, the supremum-type tests of \citet{andrews1993tests}, the multiple-break estimation of \citet{bai1998breaks,bai2003computation}, and penalised changepoint search \citep{killick2012pelt}, which let the data locate the break and report whether its location is consistent with the reference date. Equation~\eqref{eq:p2-its} is therefore the \emph{descriptive} object of the study, a convenient summary of before and after; the inferential objects are the unknown-break estimates, their confidence sets, and the placebo breaks at other dates in the baseline, as specified in Part~IV.282283The descriptive use of the date has a positive justification that goes beyond the avoidance of a fallacy. Dividing the window at a date that the whole community recognises makes the comparison between periods legible, and legibility is a scientific virtue when the quantities compared are themselves complicated. The baseline of 2018--2022 is long enough to contain the whole transformer era and the second half of the deep-learning wave, so that the ``before'' is not an AI-free past but a period of fast change in its own right; the research period of 2022--2026 is long enough for publication lags, which in computer science run from a few weeks for a preprint to more than a year for a journal, to have worked through for all but the last months. Any claim of the form ``after November 2022, $X$ increased'' in this manuscript is to be read as a comparison of two periods so defined, with the structural-break analysis as a separate and more demanding question.284285One further confounder is specific to the present study and is treated in the next section. If language models changed how papers are \emph{written}, then every text-based measure, from keyword-driven classification to embedding-based topic detection, changed its meaning at approximately the reference date, and a break in such a measure may reflect a change in prose rather than in research.286287\section{What the literature already claims about the post-2022 period}288\label{sec:p2-claims}289290A substantial literature has formed since 2023 on the consequences of large language models for science and for computer science. Its findings are the prior against which the corpus-level results of Part~IX will be compared, and its limitations are the reason the corpus-level test is needed. Five claims recur.291292\subsection{The growth of AI research}293\label{sec:p2-claim-growth}294295That AI research is growing faster than other fields was established before ChatGPT. \citet{krenn2023forecasting} built a semantic network from the AI and machine-learning literature of the preceding three decades, found that the number of papers in it had been growing exponentially with a doubling time of about two years, and used the network to forecast which concept pairs would be studied next. The Stanford AI Index reports, which compile counts from a range of sources, report continued growth in AI publications and in the number of notable models, with the share of such models produced by industry rising to a large majority by 2023 and 2024 \citep{maslej2024aiindex,maslej2025aiindex}; the reports are compilations rather than primary studies, and their source definitions change from edition to edition, so their numbers are not comparable with those of this study and are cited here for the direction of the claim only. The arXiv's own monthly statistics show that computer-science submissions overtook those of every other archive during the 2010s and continued to accelerate after 2022 \citep{arxiv2026stats}. Industry's growing weight in the field, documented by \citet{ahmed2023industry} for the period up to 2022, is the component of this claim that the present study can test most directly, through the affiliation data of the corpus (H6).296297What this literature does not establish is whether the growth after 2022 differs from the growth before it, either in rate or in kind, because the studies either end in 2022 or do not model the pre-period trend. H1 is the corpus-level version of the question.298299\subsection{LLM-assisted writing}300\label{sec:p2-claim-writing}301302The most direct evidence of a change in the literature after 2022 is a change in its language. \citet{liang2024mapping} estimated, with a distributional method that measures the frequency of words whose use the models favour, that the fraction of abstracts in computer science with substantial LLM-modified content rose rapidly from the end of 2022, with computer science leading other fields; the same group found a comparable rise in the reviews of major machine-learning conferences \citep{liang2024monitoring}, and later extended the method to writing across society \citep{liang2025widespread}. \citet{kobak2025delving} obtained the same conclusion for biomedical abstracts by a different route, measuring the excess frequency of words such as ``delves'' and ``underscores'' relative to a pre-2023 baseline, in a study designed to need no training data from the models themselves; earlier and smaller analyses of the same marker words had reached the same result \citep{geng2024writing,gray2024contamination}. Surveys of researchers confirm the adoption that the text analyses infer: a Nature survey of 1{,}600 scientists found wide use of generative tools for writing and coding, with strong concerns about reliability and misuse \citep{vannoorden2023survey}, and the multidisciplinary collection assembled by \citet{dwivedi2023sowhat} and the integrity analysis of \citet{binnashwan2023chatgpt} document the reactions of the research and education communities in the first months.303304For this study the finding is a measurement problem before it is a result. The classification of Section~\ref{sec:p2-ontology} and the embedding methods of Part~IV both read text, and text that is partly produced by a model trained on the pre-2023 literature is, in a precise sense, less informative about what is new in a paper. Two responses are built into the design. The venue and arXiv signals of equation~\eqref{eq:p2-score} are immune to prose; and the robustness checks of Appendix~D re-estimate the main breaks with classifications that use no keyword signals, so that a break present in the keyword-based series and absent in the venue-based one is flagged as possibly linguistic.305306\subsection{Illusions of understanding and the epistemology of AI in science}307\label{sec:p2-claim-epistemic}308309A second body of work concerns not the quantity of research but its quality. \citet{messeri2024illusions} argued that the uses of AI in science that researchers find most attractive, as oracle, surrogate, quant and arbiter, each create an illusion of understanding, and that the resulting monoculture of methods and of questions is a risk to the scientific enterprise rather than merely to individual studies. \citet{birhane2023science} raised the related concern that language models, used as instruments, embed the biases and the errors of their training corpora into the record. These arguments are theoretical and, so far, untested against the publication record; the knowledge-graph analyses of Part~VIII, which track methods and datasets across papers, are the closest this study comes to a test of the monoculture claim.310311\subsection{Benchmark saturation and the reproducibility of evaluation}312\label{sec:p2-claim-benchmarks}313314The benchmark culture of Section~\ref{sec:p2-dl} produced a literature on its own dynamics. \citet{ott2022benchmarks} mapped the creation and saturation of benchmarks across AI tasks and found that many become saturated within a few years of their creation while others are abandoned before saturation; the critiques of \citet{dehghani2021lottery} and \citet{raji2021benchmark} argued that the choice of benchmark is itself an under-examined research decision and that no benchmark can measure general capability. The post-2022 literature extended the concern to \emph{contamination}, the presence of test data in the training corpora of models trained on the web, which makes a benchmark result uninterpretable without knowledge of the training data that the model reports of Table~\ref{tab:p2-timeline} typically withhold. \citet{kapoor2023leakage} documented data leakage as a general cause of irreproducible results in machine-learning-based science, and the agenda of \citet{hendrycks2021unsolved} placed robustness, monitoring and alignment among the open problems of the field. RQ7, which asks whether the period's advances are breakthroughs or incremental benchmark gains, and RQ9, on reproducibility and evaluation standards, are the corpus-level counterparts of these claims.315316\subsection{Concentration, disruption and the pace of science}317\label{sec:p2-claim-pace}318319The last claim is the broadest. \citet{park2023disruptive} found, with a citation-based disruption index applied to papers and patents over six decades, that work has become less disruptive and more consolidating on average across fields; \citet{chu2021slowed} found that in large and fast-growing fields, canonical papers become entrenched, new ideas are less likely to displace them, and progress in the sense of turnover at the top slows. Both results predict that a field growing as fast as AI after 2022 should show rising concentration of citations on a few canonical works and a lower rate of disruption, independent of any product release. Hypothesis H5, on citation inequality, is formulated with that prediction in mind, and the difference between the post-2022 period and the baseline is the quantity of interest, not the level.320321\subsection{Why the claims need a corpus-level test}322\label{sec:p2-why}323324The studies reviewed above share four limitations that the design of this paper is intended to remove. They are, almost without exception, studies of AI alone, so that they cannot say whether the growth they document came at the expense of, or in addition to, the rest of computer science; the present corpus covers all forty categories with the same instrument. They rely on a single source (arXiv, a conference's reviews, a database of abstracts), so that coverage effects of the kind described in Section~\ref{sec:p2-disagree} cannot be separated from research effects; the present corpus links four sources and reports results by source. They use vocabulary as evidence, which, as Section~\ref{sec:p2-claim-writing} shows, is the quantity most directly altered by the models; the present design includes classifications that use no vocabulary. And they date the change to the release of ChatGPT without testing the alternative dates of Table~\ref{tab:p2-timeline} or the placebo dates of the baseline; the present design does both.325326None of this guarantees that the corpus-level test will overturn the claims; it may confirm them. Its contribution is to subject them, for the first time at the scale of the whole discipline and with a preregistered protocol, to the conditions under which a claim of structural change can be distinguished from the ordinary growth of a large field. Part~III describes the corpus on which that test is run.327328