An early post asking what it means when machines do things we cannot fully explain. Written before generative AI was a household concern, but the 'ghost in the machine' metaphor already does important work here.
How to use this guide. Posts are grouped by theme rather than chronology. Within each section, entries are ordered to build on one another where possible. Each post stands alone, but readers interested in the core metaphor thread are encouraged to start with Section 2 and read it through. Posts marked guest were written by invited contributors. Posts marked series form a trilogy or sequence. Annual round-ups are marked round-up.
Section One
Origins and Early Encounters
The blog's engagement with AI did not begin in 2023 with ChatGPT. These posts show a longer view — from early algorithmic unease in 2016 through to the first encounter with generative AI and the curiosity it sparked.
-
May 2016
-
Sep 2018Drawing on experience with nanotechnology and synthetic biology engagement, this post asks whether we can anticipate how publics will react to AI and what lessons earlier controversies offer.
-
Sep 2020This post deals with early anxieties about automating decision-making in the grading of A-level assessments, exacerbated by Boris Johnson's use of the phrase 'mutant algorithm'.
-
Dec 2021Prompted by a wave of AI-generated artwork, this post begins asking what AI means for how science is communicated and understood.
-
Jan 2023The blog's first post in the ChatGPT era. It asks what the new wave of generative AI might mean for education, communication and daily intellectual life, with early reflections on the dangers of 'knowledge pollution'.
-
Feb 2025A reflective post on why metaphor-watching has become a personal as well as intellectual refuge, and how AI became central to that project.
Section Two
Metaphors: The Core Thread
The core of the blog's AI coverage: a sustained, evolving investigation into how metaphors shape what we think AI is, can do, and should do. Posts are sorted thematically then chronologically. The three sub-themes below are not watertight — many posts straddle more than one — but they offer a useful orientation.
Metaphors we use to think about AI
These posts study the metaphors that circulate in public and media discourse — in journalism, social media, general commentary — and shape how non-specialists imagine what AI is and what it means.
-
Apr 2024A systematic foray into the metaphor landscape: collecting, sorting and analysing the metaphors that writers, technologists and commentators use to make AI legible.
-
Aug 2024As AI discourse intensified, so did its metaphor density. This post surveys the proliferation and asks: what does it mean when a technology attracts so many different imaginative frameworks at once?
-
Jan 2025A thematic study of food and digestion as metaphors for AI — training data as food, models as consumers, outputs as produce, and more.
-
Jan 2026A close analysis of 'slop', the word that crystallised anxieties about low-quality AI-generated content, and its remarkable semantic fertility. Companion to the enshittification post in Section 3.
-
Feb 2026A reflection on the epistemological role of metaphor itself — nets cast over AI to produce provisional understanding — and what happens when they fray.
-
Feb 2026Revisiting a classic paper by Noel and Amanda Sharkey on AI and natural magic, showing how AI has always invited magical thinking, and what it costs us when we indulge it.
-
Apr 2026Prompted by John Naughton's Observer piece on ten AI metaphors, this post examines whether 'cultural technology' — a more neutral, anthropological framing — offers something useful.
-
Apr 2026Tracing the genealogy of a single viral metaphor, 'Habsburg AI', and what it reveals about how critical AI discourse is conducted.
-
Nov 2025A short overview and entry point for the trilogy below.
-
Nov 2025A programmatic statement about why metaphor-watching needs to become systematic, collective and institutionalised. Part one of a research trilogy.
-
Nov 2025A survey of the academic landscape: what topics scholars were working on with regard to AI metaphors (up to 2025), and where the gaps remain. Part two of the trilogy.
-
Nov 2025Tracking how the metaphors used for AI have shifted since 2022 and arguing for the significance of those shifts for public understanding. Part three of the trilogy.
-
Apr 2026By Barbara Gabriella Renzi and Giulio Napolitano. A comparative study showing that AI metaphors are not universal and that language and culture shape the imaginative frameworks available to us.
Metaphors we use to build AI
Metaphors that originate inside the technical practice of AI — used by researchers, engineers and developers to describe, design and reason about AI systems. These often cross over into public discourse, carrying their technical assumptions with them.
-
Apr 2024Tracing the emergence of 'model collapse' and related contamination metaphors — showing how a technical concept becomes culturally loaded through the language used to describe it.
-
Jun 2024A technical look at how metaphors have been identified, from manual to automatic analysis, and from Lakoff and Johnson in the 1980s to the use of LLMs via natural language processing and computational linguistics.
-
Mar 2026Not just what metaphors humans use for AI, but what metaphors AI models deploy internally. An intriguing look at the metaphorical furniture built into the systems themselves.
-
Apr 2026Two field-internal metaphors examined for what they reveal about how AI developers frame safety and strategy: 'sandbox' (a safe testing environment) and 'moat' (a competitive advantage).
-
Apr 2026Sparked by a viral tweet claiming that 'DNA is 4 billion year old, completely undocumented, vibe-coded spaghetti', this post unpacks how AI developer vocabulary is being borrowed to re-describe biology itself. (Links to Section 4)
Thinking with AIs about metaphors
A distinctive strand in which AI systems become active participants in metaphor analysis — not just objects of analysis but interlocutors. These posts raise interesting questions about what it means to ask a machine about its own language.
-
Mar 2023An early (and quite amusing) chat with ChatGPT about the nature, creation and processing of metaphors, including reflections on pattern recognition.
-
Oct 2023A discussion with ChatGPT about the metaphors it would use to describe itself — which revealed a striking tendency towards magical metaphors.
-
Jul 2024Using Claude to explore how machine metaphors operate in biology, asking what an AI makes of the metaphors used to describe processes that AI itself is now being used to model. (See also Section 4)
-
Feb 2026A snapshot of the metaphor explosion that greeted Moltbook, a social network for AIs — capturing in miniature how quickly and wildly metaphors multiply around AI novelties, with AI as both subject and occasional commentator.
Section Three
Language, Words and Meaning
Beyond metaphor, these posts dig into specific words, linguistic phenomena, and the relationship between how we talk about AI and how we think about it. Polysemy, prompts, neologisms and the politics of naming.
-
Jan 2023A dissection of the quite unique metaphor of 'common sense as the dark matter of AI', including a discussion with ChatGPT about that metaphor (and a mistake it makes).
-
Oct 2023Around the first AI safety summit there was a lot of talk of 'frontier' models. But what does that word actually mean and imply for AI risks and safety?
-
Nov 2023A collaboration with Alan Miguel Valdez exploring the concept of 'frontier' further and linking it to the myth of the US frontier.
-
Feb 2024What happens to truth and the language of truth when AI can generate plausible-sounding falsehoods at scale? An examination of a vocabulary under pressure.
-
Jun 2024Drawing on J. L. Austin's speech act theory, this post asks what prompts actually do and why the language of 'magic words' is both apt and revealing about how we imagine our relationship with AI.
-
Sep 2024A meditation on the word 'intelligence' itself, whose endless meanings make precise definition both necessary and elusive.
-
Aug 2025The curious career of 'vibes', from spiritual informality to 'vibe-coding', used as a case study to trace how vernacular words get drafted into technical AI discourse.
-
Oct 2025A linguistic autopsy of one of the era's most distinctive coinages — tracing its origin, spread and meaning, and what it tells us about collective anxieties around platform and AI degradation. (See also 'slop' in Section 2)
-
Jan 2026The emergence of 'AI vegan' — meaning someone who refuses AI-generated content on ethical grounds — as a window into how moral identities are being constructed around AI choices.
-
May 2026A companion to the post on truth and post-truth, tracing the word 'fake' from its murky origins in 18th-century criminal slang, through 'fake news' and 'deepfakes', to the current crisis of reality in the age of generative AI.
-
May 2026Introducing a key paper on 'strategic polysemy' — the deliberate exploitation of multiple meanings in AI discourse for purposes of hype and power. Part one of a sequence.
-
May 2026Extending the polysemy analysis to AI safety discourse, showing how the same word can mean radically different things to different actors, with real consequences for governance. Part two.
-
May 2026Making the case that polysemy literacy — awareness of how multiple meanings of AI concepts emerge and are used — should be a core component of AI literacy education. Part three.
-
Jun 2026The final post in the polysemy series, going back to the 1970s to explore how an expert in the study of social interaction foresaw the pitfalls of AI systems. Part four.
Section Four
AI Meets Biology
A thread drawing on long experience in studying the language of genetics and genomics. What happens when the vocabularies of molecular biology and AI begin to borrow from each other?
-
Oct 2023A first exploration of the convergence between DNA-as-language metaphors and LLM terminology — both intellectually productive and conceptually treacherous.
-
Jul 2024Inspired by Benjamin Labatut's novels on mathematics and science, this post asks deep questions about meaning in LLMs and whether mathematical structure can bear the weight we place on it.
-
Aug 2024As researchers began applying LLM architectures to genomic sequences, the metaphor of 'DNA language models' raised profound questions about meaning, translation and what it means to 'understand' a genome.
-
Jan 2025By Christian Gude. A synthetic biologist asks how AI is changing the field's self-understanding and whether the controlling metaphors of synthetic biology are shifting as a result.
-
Dec 2024A playful post on chatting with a cockroach through the medium of AI at the Cambridge Museum of Zoology, and the implications this may have for science communication.
-
Mar 2026A synthesis comparing the history of metaphor in genetics (blueprints, codes, programmes) with the current metaphor ecology of AI (parrots, assistants, minds) to draw lessons about communication and risk.
Section Five
AI in Society and Culture
How AI is being absorbed into culture, identity and everyday life — from the social dynamics of AI-model naming to what AI chatbots do to our sense of relationship and truth.
-
Apr 2023The naming of AI models as animals — llamas, alpacas, falcons — as a cultural phenomenon. What does this metaphor family tell us about how the field sees itself?
-
Aug 2023An experimental and speculative post drawing an analogy between the political polarisation of GM crops and the emerging cultural divisions around AI.
-
Dec 2023A personal and methodological reflection linking earlier work on climate change communication to the new project on AI, showing the continuity of concerns beneath different subject matters.
-
Jan 2024An analysis of the double movement in AI discourse: as we project human qualities onto machines, we risk diminishing what is genuinely human.
-
Oct 2024Tracing 'superintelligence' from its theological roots to Sam Altman's recent pronouncements, showing how this concept carries far older anxieties about the creation of something beyond our control.
-
Oct 2024Reflecting on the word 'playing' — how the playfulness of early AI interactions conceals real dangers, and how the experimental framing both enables and disguises what we are doing.
-
Nov 2024An experiment in using AI to discuss climate change, asking what AI-mediated conversation reveals about how we think and feel about a vast and difficult subject.
-
Sep 2025Engaging with a complex paper on 'computational hermeneutics' (and with Gadamer and Dilthey), this philosophical post asks whether interpretation and understanding are things AI can genuinely do.
-
Nov 2025By Andrew Maynard. On 'parasocial' becoming Cambridge Dictionary's 2025 Word of the Year, and what it means that our growing attachments to AI chatbots have a name.
Section Six
Ethics, Safety and Governance
How AI risk, responsibility and regulation are being imagined, argued over and framed. These posts engage with the policy and ethics debates directly, through the lens of language and public understanding.
-
May 2023By Dimitrinka Atanasova. An early examination of whether AI tools help or harm researchers from non-Anglophone backgrounds — a practical equity question often absent from mainstream AI ethics discourse.
-
Jun 2023A critical examination of the 'existential risk' framing: how the term functions rhetorically, what it closes down as well as opens up, and the journey from alarm to the more procedural language of 'alignment'.
-
Nov 2023By Alan Miguel Valdez. Against the backdrop of the Bletchley Park AI Safety Summit, this post brings AI back to the pavement level — literal robots, actual communities, and what governance means on the ground.
-
Feb 2024On Goody-2, the satirical AI that refused every request on safety grounds — a humorous intervention in the responsible-AI debate that reveals its contradictions from the inside.
-
May 2024Surveying the rapid proliferation of 'AI safety' as a term and asking — quite urgently — what it actually means across its very different uses.
-
Jun 2025Reporting on a social-science webinar on what AI researchers say about public reception of AI, examining the tensions between institutional engagement rhetoric and actual practice.
-
Aug 2025Looking at the history of AI hype cycles through two salient metaphors — 'winter' and 'bubble' — and asking what they tell us about whether the current moment is different, or the same old story.
-
Sep 2025Engaging with Alondra Nelson's proposal to apply the genomics ELSI framework to AI, reflecting on technology governance, engagement and public participation.
Section Seven
Annual Overviews and Reflections
Each year the blog produces a round-up of all posts written during that year. Together they form a record of how AI (and the blog's engagement with it) has evolved.
-
Dec 2023The year AI became something we could all use and were all talking about. A useful map of the blog's first intensive year of AI coverage.
-
Dec 2024A year in which the metaphor project deepened and the AI-and-biology thread emerged strongly. Also covers non-AI posts for readers interested in the broader scope of the blog.
-
Dec 2025The year of the metaphor observatory trilogy, enshittification, vibes, AI veganism, and a blog migration.