While robot pets are not likely to become the norm, various forms of AI technology continue to make their way into veterinary practices. The decrease in cost and increase in ubiquity of computing capacity has led to increasingly powerful software that no longer simply reduces tedious work (e.g. accounting, word processing) or is an “expert system” that provides higher level technical capabilities, but, rather, to machine learning (ML). ML consists of algorithms that improve with exposure to data, analyzed using complex statistics and probability rather than the absolutism of the binary system we typically associate with computers. The result of ML is software able to rival or exceed human capabilities in, currently, very targeted applications. This type of focused ML is also known as “narrow” artificial intelligence (AI). We are likely decades away from the “technological singularity” of science fiction and futurists’ dreams where a “general” artificial intelligence develops and surpasses overall human intelligence. However, in the estimated few decades before humans are no longer the most intelligent being in the known universe, “narrow” artificial intelligence can provide a significant boost to all aspects of human endeavors, including veterinary medicine. The following discusses leading-edge resources and the future that is coming soon to your practice. Time, time, time Veterinarians’ most limited resource is time. As such, tools capable of saving time are highly valued in practice. There are very limited “expert” software systems in use in veterinary medicine, but these can save significant time. For example, software developed by this article’s co-author, Dr. Delaney, generates custom pet food recipes daily, which previously would have taken 7.6 years for a board-certified veterinary nutritionist using previous software to create. ML offers even greater opportunity. While the already available ML-powered offerings for veterinarians are time-savers, they are not yet improvements on what humans can achieve without their use. This likely will change based on recent human medicine studies.1 Similar image recognition ML can be applied to microscopy as seen with the SediVue Dx urine sediment and ProCyte One hematology analyzers. ML has also now started to give new insight into other diagnostic test data beyond images, as seen with the University of California, Davis’s (UC Davis’s) Addison’s screening AI, as well as RenalTech, which relies on existing biochemistry panel data to flag or identify potential disease earlier and/or with increased confidence. With increased access to “big data,” especially from centralized veterinary medical record databases, ML use will further grow and prove itself in other areas as it can “train” more and improve its algorithms. Enhanced cognition Soon on the horizon, there will be more hybrid systems that mix “expert systems” or different ML solutions. Efforts to leverage ML-powered natural language processing (NLP) in veterinary medicine are already underway with services like Talkatoo, Dragon Veterinary, and Talkingvet. These automated dictation or speech to text services use technology like that found in many smart devices, including Google Assistant, Amazon’s Alexa, and Apple’s Siri. Similarly, medical transcription software could convert or transcribe spoken words, which are then reviewed by another ML system that uses those words to infer diagnoses. Thus, a system able to scan veterinary medical records to suggest potential diagnoses,2 coupled with an NLP dictation solution like those already on the market could, theoretically, not just save typing time, but also digitally enhance the veterinarian’s cognition. Garbage in, garbage out Using image recognition as another example, one can imagine a ML system combining digital radiographs and digital microscopic images of an aspirate and “expert system” logic to then suggest additional diagnostics, referral, prognoses, and/or treatment options. These systems, by design, would become more robust as more tagged data by subject matter experts/specialists are made available. An image recognition system trained on a vast library of accurately read radiographs will perform better than one exposed to limited and/or badly labeled radiographs. Many of us have already helped create good training images (often unknowingly). For example, to help Google train its own systems, the search engine uses a service called “reCAPTCHA” and leverages many “volunteers” to identify objects within an image to prove they are not a “robot” to websites. These human-tagged images can then be used as training data to improve Google’s image recognition. As veterinarians continue to use more cloud-based or connected ML-powered tools, they will create more high-quality data able to be used to further train the ML-powered tool. These services will then improve, leading to a positive feedback loop where better ML-powered tools result in increased adoption and use, which then generates more training data, which, again, improves ML-powered tools. What to expect For physicians and veterinarians practicing today, highly transformative technology will likely occur during many of our medical careers. Today, we are starting to see the first veterinary offerings to combine these emerging cognitive technologies. . Additionally, the authors of this article recently worked with a team of developers and veterinary nurses to develop a technology combining NLP and a user customizable expert system to improve client communications and patient outcomes through automated follow-up. The goal of systems like these and others is to capture back time for other “human-only” tasks, while improving on what could be achieved by a human alone. In the coming years as veterinary medical professionals, we should expect to see more such offerings that will assist with diagnostic imaging, microscopy, differential diagnoses, medical records, and client communications. They should, collectively, save time, increase the speed of diagnosis, provide additional peace of mind, offer improved client experience and, most importantly, improve patient outcomes. Today’s veterinarians will, once again, need to proactively adapt. We should evolve to take full advantage of all technological resources to improve patient care and make it more efficient—until decades from now, when we might actually say those famous words of Ken Jennings, the Jeopardy! champion who lost to IBM’s Watson, “...welcome our new computer overlords.” Sean Delaney, BS, DVM, MS, DACVIM (Nutrition), is the founder of Vet Assistant AI, Inc., (VAAI) in Northern California. Yuki Okada, BA, DVM, PhD, is the owner of Okada Ventures, Inc., and co-owner of VAAI and Seven Hills Veterinary Hospital, a general practice in San Francisco. References Bai et al. Radiology. 2020 Sep; 296(3):E156-E165. Nie et al. NPJ Digit Med. 2018 Oct 24;1:60. 3. Jumper et al. Nature. 2021 Aug; 596(7873):583-589.