The concept of “artificial intelligence” has been in the social consciousness for over 200 years. Cautionary tales about the consequences of humans creating artificial life have been a staple in literature and film since Mary Shelley wrote Frankenstein in 1818, in which a fabricated "monster" with an inherently peaceful nature is ultimately driven to violence after being repeatedly rejected by humans (including his own creator). The term “robot” first entered the lexicon in the 1920 play R.U.R. by Karel Čapek, in which artificial workers rebel against their masters and bring about the extinction of the human race. (The word “robot” comes from Czeck robota, “forced labor”.) This theme² has been repeated in numerous sci-fi literary and cinematic classics that people have consumed for generations, myself included; I’m of the Terminator generation.

AI in our non-fictional lives is a relatively recent phenomenon. In November 2022, the private company OpenAI launched ChatGPT, its AI chatbot, and two months later it already had 100 million active users. AI-generated art, uncanny “deepfakes”, and other AI-generated videos almost indistinguishable from real footage are now ubiquitous on social media. In late June, a new band appeared on Spotify and garnered a sizeable following in a matter of days. The band? Entirely AI generated. Many apps and other tech we use every day (like iNaturalist, Microsoft Teams, Google, and our cars!) use some kind of computer-aided algorithm to assist our work in various ways. We even have slurs for AI bots now – "clanker" being one (but as a good friend said, I'm pretty annoyed that "toaster" isn't.

The answer is unequivocally yes. Within the past few years, the Florida Museum of Natural History recently filled among the first curatorial positions for AI in the fields of Natural History and Archaeology, where AI is being used to inform how organisms evolve by aggregating the data that would take a human two hundred 8-hour days to collect in a single hour – saving staff time and institution funding.

AI is being used to benefit wildlife conservation by analyzing mountains of data in a fraction of the time it would take humans to do. AI-assisted game cameras are being used to alert local farmers of the presence of tigers in India, and a researcher at the University of Alberta is using AI to analyze thousands of hours of recordings of nighthawk calls to learn how and when they use their foraging vs. nesting habitats (see this article from Yale).

There are also potential and proposed ideas for how AI could be used to benefit Natural History collections and the curators who care for them.

Detailed habitat descriptions often accompany natural history specimens. These descriptions can include information about geology, hydrology, and other species found at the place of collection. These descriptions could be analyzed with AI tools (such as Machine Learning, a subcategory of AI that uses algorithms to find patterns in data) and used to train predictive models with georeferenced records. These models could reconstruct the historic range of species, predict where species might be found but have not been documented, update species range maps, and be used to document the spread of invasive species (see this article in Trends in Ecology & Evolution). Machine Learning could also be used to accelerate the analysis of mountains of data to predict the conservation status and extinction risk of rare and endangered species and update their listing on the IUCN Red List in a fraction of the time humans can do it (see this post from the Natural History Museum in London.)

The above are only a few examples, but there are many others. I invite the reader to investigate further!

In other AI-adjacent news, biotechnology company Colossal BioSciences announced in April 2025 they had brought the Direwolf back from extinction, and they used AI to do it. More recently, they announced their intention to bring the South Island Giant Moa (Dinornis robustus) back from extinction as well – only one of many species of moa, extinct flightless birds endemic to New Zealand. This goal is orders of magnitude more ambitious and difficult than what they accomplished with the wolves (and likely impossible). The announcement was met with skepticism and opposition by ecologists, geneticists, and Indigenous groups. Peter Jackson, The Lord of the Rings motion picture trilogy director (and keeper of a large private collection of moa specimens) is helping to fund the endeavor.

Whether de-extincting the moa would count as a “benefit” to Natural History remains to be seen – but if it happens, AI will have made it possible.

Additional photo details: This Upland Moa (Megalapteryx didinus) is one of the coolest existing specimens of moa (in my opinion) but is NOT the species Colossal Biosciences wants to “de-extinct”. A-B, Head and neck from left (A) and right (B) side. C-D, Right lower leg in medial (C) and lateral view (D).

A lot of AI-generated material on social media is “clickbait”, provocative content that with little substance that is made to drive post engagement, or “ragebait”, inflammatory content that often stokes political division (which also drives post engagement). ChatGPT is being used for a variety of tasks that are questionable at best – the reader is probably already aware of a few examples, and likely has already used ChatGPT for one reason or another. AI-generated images of fictional plants and animals are shared on social media with identifications of actual species (my personal pet peeve), which I feel only widens the separation of people from nature. The internet is now rife with websites constructed with AI writing tools that contain inaccurate information about the natural world. Search engines will prioritize information from the most popular websites, even if they are AI-generated and incorrect, making the task of searching for factual information even more difficult. Foraging for wild foods has recently become popular, but the information regarding edibility of wild plants and fungi supplied by a search engine can be dangerous if it’s not correct. ChatGPT is being used to trawl the internet and churn out books for sale on Amazon written by fabricated authors that are full of false information.

Water and energy use by data centers have been a concern for several years, particularly when they are built in already water-stressed areas; the use of AI tools is poised to exacerbate these existing issues. Rather than summarize these complex issues here, I recommend readers read this article from Yale and this article from MIT. The short version is that the alarm has sounded on water and energy use by AI, but at the time of this posting, the metrics being reported are extrapolations – but that is largely because either actual measurements are not being taken, or the data is not being disclosed². In any case, energy use and water use by AI are serious concerns that need to be addressed.

When I use iNaturalist’s AI-assisted identifications, it usually offers several options with a “% match” alongside. Depending on the quality of the photo, iNaturalist’s top suggestion may not be correct, but it is usually one of the alternatives provided. I cross-check its conclusions by looking up the scientific names in the appropriate field guides or online resources (or, since I have the privilege of stewarding a natural history collection, I can check real specimens), and then I compare field marks to the photo I’m trying to identify. Although iNaturalist takes range into account when making IDs, I also consult range maps. In some cases I may consult colleagues or other experts. Finally, once I decide on an ID and upload my observation, the human iNaturalist community checks it, often offering comments to support their argument if they disagree. In fact, having humans check and confirm IDs helps improve the accuracy of the identification models.

1. This preprint also reported that the people in the non AI-assisted group for their writing task showed significant increases in brain connectivity and had better recall after they were switched to the ChatGPT-assisted group. This seems to suggest AI may boost human cognition if it’s used as a post-hoc tool, rather than a load-bearing support beam. However, this study hasn’t been peer reviewed and had a small sample size, so I would like to see it repeated with more people!
2. To be fair, there have been plenty of stories about "good robots", too: Wall-E, Short Circuit, Data of Star Trek (but not his "evil" brother Lore!), and many droids from the Star Wars universe come to mind; and even in Terminator 2: Judgement Day, the "bad robot" from the first film was reprogrammed to be a "good robot". These "good robot" stories often ask the viewer to question whether the robots are alive/sentient/deserving of full personhood – another theme that saturates the social consciousness: If we create artificial life, does it deserve rights?
3. Google has disclosed what percentage of their energy processes are used for Machine Learning (ML). One study found that the energy use impact (measured in CO₂ emission equivalents) was 300x greater for smartphones than that for ML. This was due to inefficiencies in energy use in using the smartphone’s processes for ML, vs. a centralized data center that is optimized for efficiency. The study also found that energy loss from smartphone chargers can be four times greater than the energy needed to run the smartphone itself. (Wireless chargers are particularly wasteful and inefficient.) Given the ubiquity of cell phones, this may suggest energy wastage by charging them is of greater concern than energy use by AI processes. (Others may argue it smacks of “the responsibility of environmental stewardship lies with individuals, not with corporations”.)