A report from 2035. Most of it is made up. Not all of it.
THE GULF, 2035 — The region’s last practising veterinarian retired on Tuesday, completing what officials called “a smooth transition to AI-first animal care” and what trainers called “Tuesday”.
His replacement, VetBot 7, arrived in a white van with a drone, a subscription agreement and 400 pages of terms and conditions that nobody read, including VetBot 7.
“This is a historic day for animal health,” said a spokesperson for the company, who asked not to be named because it is a language model.
A quiet takeover
The machines didn’t arrive all at once. They came in the way most technology comes into a clinic: through the paperwork.
First came the AI scribes, which listened to consultations and wrote the clinical notes. Vets loved them. For the first time in the history of the profession, the records contained complete sentences. Nobody read them, but they were there.
Then came the radiology software, which read X-rays in seconds and never needed coffee. It had learned from millions of images of dogs and cats, and it was very, very good at dogs and cats. Its first camel radiograph was reported as “a large dorsal mass, likely neoplastic; biopsy recommended”. The hump is doing well.
Camels, of course, had seen this before. Robot jockeys, some of them rebuilt from cordless drills, have ridden in camel races here since 2005, when they replaced child jockeys. “They’ve worked with robots for thirty years,” said one trainer. “They still don’t like them. Why would they like this one?”
Early results
By most measures the new system is a success.
VetBot 7 can detect lameness from a ten-second phone video. In its first week it found a grade 3 lameness in a racing camel, a grade 2 in the handler and a mild but “clinically significant” asymmetry in the trainer, who has asked for a second opinion.
Its dosing is precise to the microgram. The instructions say “administer orally”, which works well until they meet the camel.
It’s also available around the clock. Owners describe it as patient, polite and always online. “When my camel lost, it said it understood how I feel,” said one owner. “Then it reminded me that it can’t give financial advice.”
Unexpected problems
Engineers admit that some tasks have been harder than expected.
The system can recognise more than 4,000 conditions from a single photograph. It cannot yet get a camel to stand still for the photograph. Researchers call this a known limitation. Roboticists have a name for it, Moravec’s paradox: the things we find hard, like diagnosis, turn out to be easy for machines, while the things any stable hand can do, like getting a frightened animal to cooperate, turn out to be very hard. “Diagnosing a camel is basically solved,” one engineer said. “Catching one is still an open research question.”
Communication has also needed work. Most cases still begin with a WhatsApp voice note, recorded in three languages and a strong wind. VetBot 7 transcribes these perfectly, then asks the handler to confirm consent. The handler calls the manager, the manager calls the owner, and the owner, who is on holiday, replies with a thumbs-up emoji. VetBot 7 has logged this as informed consent for 214 procedures.
Then there’s the question of who is responsible when things go wrong. The terms and conditions state that the algorithm is not responsible for the diagnosis, and neither is the company. Legal experts expect responsibility to fall, as it always has, on whoever was holding the lead rope.
Regulators say they are watching closely. In the United States, veterinary devices have never needed approval before going on sale, and officials elsewhere have promised to review the matter as soon as someone explains to them what a camel is.
The human touch
Last month the company quietly hired back its first human, under the job title “biological interface specialist”. Her duties include holding the leg, drinking coffee with the owner, explaining the bill, noticing that the handler looks worried, and being blamed.
“Honestly, it’s the same job,” she said. “The notes are just longer.”
So, could it happen?
Parts of it already have. AI reads radiographs, drafts clinical notes and scores lameness from video today, and I use some of it myself. But look at what the satire leaves for the human: restraint, trust, judgement in a messy yard, and the conversation with the owner. That’s not an accident.
In 2013, two Oxford economists, Carl Frey and Michael Osborne, estimated how likely 702 jobs were to be computerised. They gave veterinarians a 3.8% chance, among the least automatable fifth of their list. The economist David Autor explains why with Polanyi’s paradox: we know more than we can tell, and what we can’t spell out as rules is hard to hand to a machine. A lot of veterinary work is exactly that kind of knowledge.
So no, I don’t think the machines will replace vets. I think they’ll replace the parts of the job that were never the point, if we let them, and leave us more time for the animal and the person holding the lead rope.
The camels still won’t like it.
References
- Frey, C. B. & Osborne, M. A. (2017). The future of employment: How susceptible are jobs to computerisation? Technological Forecasting and Social Change, 114, 254–280. doi:10.1016/j.techfore.2016.08.019. Veterinarians: probability 0.038, ranked 122nd of 702 occupations.
- Autor, D. H. (2015). Why are there still so many jobs? The history and future of workplace automation. Journal of Economic Perspectives, 29(3), 3–30. doi:10.1257/jep.29.3.3
- Moravec, H. (1988). Mind Children: The Future of Robot and Human Intelligence. Harvard University Press. hup.harvard.edu
- Al Jazeera (18 July 2005). Robots ride camels in race. aljazeera.com
- Deck, A. (March 2020). How a robotics engineer accidentally upended child labor practices in the Gulf. Rest of World. restofworld.org
- Crawford, A. (July 2015). The latest sign that the robot uprising is nigh? Camel racing. Smithsonian Magazine. smithsonianmag.com
- US Food and Drug Administration. How FDA regulates animal devices. fda.gov