Species

Podcast 2 — In conversation

Speaker A: The question of what makes a species a species is genuinely one of the most unresolved questions in biology, which is surprising given how fundamental the concept is. The biological species concept — organisms that can interbreed and produce fertile offspring — has the nice clarity of a reproductive criterion but falls apart at the edges in all sorts of ways. Horses and donkeys can interbreed to produce mules, but mules are infertile — does that make horses and donkeys the same species or not? Polar bears and grizzly bears have fertile hybrid offspring in the wild — does that make them the same species? The answer biologists give is usually no to both, but the reasoning gets complicated.

Speaker B: The extinction crisis for species is one of the defining environmental stories of this century. The current rate of species extinction is estimated at anywhere from one hundred to one thousand times the background rate — the rate at which species would normally go extinct in the absence of human impact. The International Union for Conservation of Nature maintains the Red List of Threatened Species, which is the most comprehensive catalogue of the conservation status of the world's species, ranking them from least concern through vulnerable, endangered, critically endangered, and extinct in the wild, to extinct. The Red List is a sobering document.

Speaker A: Species versus subspecies — what is the distinction?

Speaker B: A subspecies is a recognised subdivision within a species — a group that is geographically distinct and has evolved measurable differences from other populations of the same species but that can still interbreed with them if brought into contact. Tigers provide a clear example: the Amur tiger of the Russian Far East and the Bengal tiger of the Indian subcontinent are different subspecies of Panthera tigris — recognisably different in size, colouring, and habitat but belonging to the same species. A subspecies is given a third name in the binomial nomenclature system. The distinction matters enormously for conservation — deciding whether a population is a distinct subspecies affects whether it receives separate legal protection.

Speaker A: What is a common mistake about the plural of species?

Speaker B: Writing or saying specieses, which is not a word in standard English. Species is one of the Latin loan words in English that has retained its Latin plural form — the plural is species, identical to the singular. This is the same pattern as series, one series, two series, and corpus in some technical uses. The confusion arises because most English nouns pluralise by adding an S, so the temptation to add another S to species to form a plural is very strong. The correct usage — there are thousands of species of beetle in that single family — sounds exactly the same in speech, which means the error only surfaces in writing.

Speaker A: Any close synonyms?

Speaker B: Kind, type, variety, and sort are the closest in the general sense of a category or type of thing. In biological contexts, taxon is the general term for any named group at any level of classification — species is a specific rank within the taxonomic hierarchy. Breed is used for domesticated animals and refers to populations that have been deliberately selected by humans rather than evolved by natural selection — a breed is not a species in the biological sense. Strain is used for microorganisms and viruses — a strain of influenza is a variant within a species. Each of these terms is more narrowly applicable than species, which remains the fundamental and universal unit of biological classification.

Podcast 3 — Prompt engineering

Speaker B: Welcome to the workshop. Today we prompt AI using the word species — building the taxonomic classification systems, biodiversity databases, endangered species tracking tools, biological data catalogue managers, and evolutionary tree visualisers that organise, analyse, and present the richly structured world of biological classification and conservation.

Speaker A: So species tells the AI to build the biological classification systems — the taxonomic databases that store species hierarchies, the conservation status trackers that monitor endangered species, and the biodiversity analysers that measure species richness and ecological health?

Speaker B: Exactly. It signals taxonomic data management, biodiversity analytics, and conservation monitoring. Let me start with a species database.

Build a species identification and classification API: accept an image of an organism and return the likely species identification with a confidence score. For each identified species, return the full taxonomic classification — kingdom, phylum, class, order, family, genus, and species — the common name, the conservation status, the geographic range, and a brief description. Support fuzzy matching for partial or ambiguous identifications and return multiple candidate species ranked by confidence when the identification is uncertain.

Speaker B: Species identification API accepts images, returns taxonomic classification, common name, conservation status, geographic range, and description with confidence scores, and supports fuzzy matching with ranked candidate species for ambiguous identifications — the AI builds the classifier, the taxonomy retriever, the conservation status lookup, and the fuzzy match ranker.

Speaker A: Returning multiple candidate species ranked by confidence when the identification is uncertain — rather than forcing a single identification — is the epistemic honesty requirement: a photo taken at a distance or in poor light may be genuinely ambiguous between two similar species, and presenting the top three candidates with their confidence scores gives the user the information they need to investigate further rather than accepting a potentially wrong single answer.

Speaker B: Ranked candidate species for ambiguous identifications as the epistemic honesty mechanism. Now the biodiversity schema.

Design a biodiversity monitoring schema: a species table recording each species with its ID, binomial name, common names as a JSON array, taxonomic classification at each rank, IUCN conservation status, endemic regions, and discovery date. A observation_records table recording each species sighting with a species ID, observer ID, date, latitude, longitude, count, and habitat type. A protected_areas table recording each designated conservation area with its ID, name, area in square kilometres, country, and the count of species recorded within it. A species_trend_view computing the change in observation frequency for each species over the past five years, flagging declining species for conservation alert.

Speaker B: Biodiversity schema stores species taxonomy with conservation status and endemic regions, observation records with location and habitat, protected area records with species counts, and a five-year frequency trend view flagging declining species — the AI builds the four-table schema.

Speaker A: The five-year observation frequency trend — flagging species whose observation frequency is declining — provides an early conservation signal before the species reaches a formal threatened status on the IUCN Red List: a species that is declining in observation frequency but not yet classified as threatened may be in the early stages of population collapse, and the trend flag allows intervention before the formal threat threshold is reached.

Speaker B: Five-year frequency trend as the pre-Red-List early conservation signal. Next, the conservation monitoring interface.

Design a species conservation monitoring dashboard for a conservation organisation: display a world map showing the geographic distribution of monitored species as coloured dots, coded by conservation status — green for least concern through red for critically endangered. Allow clicking any dot to see a species profile with its population trend chart, threats, and active conservation programmes. Show a summary panel of the count of species in each conservation category. Include a recent sightings feed showing new observation records submitted in the past seven days. Allow filtering by taxonomy, geographic region, and conservation status.

Speaker B: Conservation dashboard shows a colour-coded world distribution map, clicked-species profiles with trend charts and programmes, a conservation category count summary, a seven-day sightings feed, and taxonomy, region, and status filters — the AI builds the map display, the species profile panel, the trend chart, the sightings feed, and the filter controls.

Speaker A: The colour coding by conservation status on the geographic map — green for least concern through red for critically endangered — gives the conservation team a spatial picture of threat distribution: a region dense with red and amber dots signals a biodiversity hotspot under severe pressure, and the geographic clustering of threatened species often reflects underlying habitat loss or other regional threats.

Speaker B: Conservation status colour mapping as the threat distribution spatial pattern revealer. On to species data analytics.

Create a species richness analyser: given a set of species observation records for a defined geographic area, calculate the species richness — the count of distinct species observed — the Shannon diversity index — which accounts for both richness and the evenness of species abundance — and the proportion of endemic species. Compare the diversity metrics across multiple areas or time periods. Identify species that are present in an area today but were absent from historical records, and species that were historically present but are no longer being recorded.

Speaker B: Species richness analyser calculates species count, Shannon diversity index, and endemic species proportion, compares metrics across areas and time periods, and identifies newly arrived and historically lost species — the AI builds the richness calculator, the Shannon index calculator, the endemism counter, the cross-area comparator, and the arrival and loss detector.

Speaker A: Detecting species that are present today but absent from historical records — alongside species that are historically recorded but no longer observed — gives a complete picture of community change: new arrivals may indicate range expansion or invasive species introduction; historical losses confirm local extinctions; together they tell the story of how the ecological community is changing over time.

Speaker B: Arrival and loss detection as the complete ecological community change story. Now something whole.

Build a biodiversity data platform: aggregate species observation data from field researchers, citizen scientists, and partner organisations. Validate and clean incoming observations. Calculate diversity and trend metrics at local, regional, and global scales. Support conservation planning by identifying priority areas for protection based on species richness, endemism, and threat status. Generate annual biodiversity state reports for government and NGO stakeholders. Provide a public-facing species discovery interface for citizen scientists to submit and explore observations.

Speaker B: Biodiversity platform aggregates and validates observations from multiple sources, calculates multi-scale diversity metrics, identifies conservation priority areas by richness and threat, generates annual stakeholder reports, and provides a public citizen science interface — the AI builds the full biodiversity data management and analytics platform.

Speaker A: Identifying conservation priority areas based on the combination of species richness, endemism, and threat status — rather than any single criterion — produces the most defensible conservation investment recommendations: an area that scores high on all three measures is a location where investment will protect the most unique biodiversity most urgently, and the multi-criterion prioritisation prevents single-metric distortions.

Speaker B: Multi-criterion priority scoring as the defensible conservation investment recommendation mechanism. Last one — species API integration.

Write an integration layer for biodiversity data APIs: connect to the Global Biodiversity Information Facility, the IUCN Red List API, and the Encyclopedia of Life API to retrieve species data on demand. Cache retrieved species records locally with a configurable time-to-live to avoid excessive API calls. Handle rate limits and API downtime gracefully — serve cached data when the upstream API is unavailable and queue requests for retry when rate limits are reached. Provide a unified response format normalising the different data structures from each API into a consistent species data object.

Speaker B: Biodiversity API integration layer connects to three external species data APIs, caches records with configurable TTL, handles rate limits with queued retry and falls back to cached data on API downtime, and normalises three different response formats into a consistent species object — the AI builds the multi-API connectors, the cache layer, the rate limit handler, the retry queue, and the response normaliser.

Speaker A: Normalising the different data structures from three APIs into a consistent species object is the integration abstraction that makes the consuming application resilient to upstream API changes: when one of the three APIs changes its response format, only the normalisation layer needs updating rather than every part of the application that uses species data.

Speaker B: That is the goal — one classifying word, better-structured biological data systems. Keep it in your kit.