Artificial Intelligence (AI) is profoundly affecting our lives, particularly in the United States, where it’s being widely adopted.

I decided to use AI to do a brief assessment of its potential impact on One Health. Since One Health is so broad in scope, I focused on four basic questions. I asked Chat GPT to list the pros and cons that AI presents in human medicine, veterinary medicine, antimicrobial resistance (AMR), and environmental health. I also requested Chat GPT to provide the references and URLs for its findings.

Here are the results (with some edits):

What are the pros and cons of AI in medicine?

Pros:

  • Faster and More Accurate Diagnoses: AI improves diagnostic efficiency and capability. For example, AI and machine learning are being extensively integrated into radiology, laboratory medicine, and pathology.
  • Improved Efficiency in Drug Discovery: Drug discoveries are expected to increase because AI can analyze massive datasets and identify potential drug candidates and treatment targets. Pharmaceutical companies have fully embraced AI.
  • Personalized Treatment: AI is expected to change medicine from a “one-size-fits-all” model to a personalized one because it assists clinicians to tailor treatments based on their patients’ genetics, risk factors, and medical histories. It can also track chronic conditions and identify problems before they worsen.
  • Better Monitoring and Prevention: Wearable devices and electronic health records monitored by AI can identify health risks before symptoms get worse.
  • Support for Medical Research: Vast datasets can be analyzed more efficiently using AI. Patterns in clinical trial data, new relationships between diseases and treatments, and finding potential drug candidates are all potentiated using AI.
  • Increased Access to Care: Basic health information, triage assistance, and areas with shortages of healthcare professionals could be assisted using AI.

Cons:

  • Risk of Errors: AI can produce diagnostic errors, especially when encountering cases outside of the training data. Like humans, AI is not infallible to error.
  • Bias and Fairness Issues: AI trained on unrepresentative datasets might perform less accurately for populations that were not included in the training models, creating bias in diagnosis.
  • Privacy and Security Concerns: The use of large-scale patient data raises concerns about confidentiality, cybersecurity, and data misuse.
  • Lack of Transparency: Some AI systems can function as “black boxes,” making it difficult to understand how diagnoses and recommendations are made. This lack of transparency can contribute to legal and ethical uncertainties, especially if AI contributes to a medical error.
  • Overreliance on Technology: An overreliance on AI could reduce independent clinical expertise and judgement. Medical education might be adversely impacted if students depend too much on AI for developing critical thinking skills.
  • Legal and Ethical Challenges: Safety concerns regarding AI-related device failures and recalls could jeopardize patient care, highlighting a need for regulatory oversight.
  • Cost and Implementation Difficulties: Developing, integrating, and validating AI systems into hospitals and other healthcare settings can be expensive and technically challenging. Staff would need to be trained, and maintenance of the system would require continuous monitoring.

What are the pros and cons of AI in veterinary medicine?

Pros:

Cons:

  • Limited Veterinary Datasets: Compared to human medicine, veterinary medical datasets are more limited, reducing AI reliability.
  • Species Diversity Challenges: Since veterinary medicine takes care of more than one species, AI systems must account for them making model development more challenging, potentially impacting accuracy and validity.
  • Accuracy Concerns: Veterinary professionals cite reliability and accuracy as major barriers to adoption.
  • Hallucinations and Misinformation: (Note: Not clear what Chat GPT meant by “Hallucinations.”) Generative AI tools can produce plausible but incorrect clinical information.
  • Ethical and Liability Issues: There are ethical and liability concerns like in human medicine.
  • Cost and Implementation Barriers: Smaller veterinary clinics might face financial and technical challenges when trying to adopt AI technologies.

Table 1: Summary of AI in Human Medicine vs. Veterinary Medicine

AspectHuman MedicineVeterinary Medicine
Data AvailabilityVery large datasetsSmaller, more fragmented datasets
RegulationExtensive regulatory oversightLess developed oversight
Diagnostic ToolsWidely deployed AI-enabled devicesGrowing but less mature adoption
Main BenefitsDiagnosis, treatment, planning, monitoringDiagnostics, animal welfare, herd management
Major ConcernsBias, privacy, liability, safetyData limitations, species variation, reliability

What are the pros and cons of AI in AMR?

Pros:

  • Faster Detection of AMR: AI can analyze clinical, genomic, and laboratory data faster than traditional methods resulting in earlier diagnosis of AMR. Faster treatment decisions should help to reduce the spread of AMR.
  • Improved Antimicrobial Stewardship: Improved antimicrobial stewardship based on the AI-based clinical decision-making, reducing inappropriate antimicrobial use. This should lower the risk of AMR spreading and improve patient outcomes.
  • Enhanced Surveillance and Outbreak Detection: Large datasets from hospitals, laboratories, public health agencies, and genomic databases can be processed and analyzed in real-time using AI to improve early warning systems and better public health responses for
  • Accelerated Drug Discovery: Faster identification of novel antibiotics that conventional methods might miss also helps to reduce costs. For example, MIT researchers used AI to identify the antibiotic “halicin.”
  • Prediction of Resistance Mechanisms: AI analysis of genomic and proteomic data can predict how microorganisms develop resistance, producing a better understanding of AMR evolution and improving the targeting of interventions.
  • Optimization of Infection Control: AI can identify patients at high risk for AMR infections and recommend preventive measures, reducing hospital-acquired infections and improving patient safety.
  • Support for Resource-Limited Settings: AI-based diagnostic tools may help healthcare systems with limited specialist expertise, increasing access to diagnostic support, faster treatment decisions, and improved healthcare equity.

Cons:

  • Dependence of High-Quality Data: AI systems require large, representative datasets for training and validation. Missing or inaccurate data, inconsistent laboratory standards, or poor-quality data can produce unreliable predictions.
  • Algorithmic Bias: Most AMR datasets are developed in high-income countries with large healthcare institutions. Potential health disparities exist if datasets come from poor, underrepresented populations, creating geographic bias.
  • Lack of Explainability: Advanced machine-learning models can be seen as “black boxes,” generating mistrust in clinicians, especially if they find it difficult to understand the recommendations.
  • Risk of Incorrect Recommendations: AI models are not infallible and could lead to potential patient harm through inappropriate antimicrobial recommendation, delayed therapy, and treatment failure. All AI recommendations should undergo careful clinician review.
  • Privacy and Security Concerns: Since AI systems rely on large datasets such as electronic medical records (EMR), laboratory databases, genomic information databases, they are at risk for data breaches, unauthorized access, creating patient privacy concerns.
  • High Implementation Costs: AI requires computing resources, digital infrastructure, staff training, and ongoing maintenance, creating barriers for smaller healthcare systems.
  • Regulatory and Legal Challenges: AI in healthcare requires clinical validation standards, regulatory oversight, and liability protections when AI contributes to medical errors.
  • Overreliance on Technology: AI should supplement, not replace, microbiologists, infectious disease specialists, pharmacists, and clinicians. Excessive dependence on AI may reduce critical clinical evaluation and judgement skills.

Table 2: Summary of AI in AMR

ProsCons
Faster resistance detectionRequires high-quality data
Improved antibiotic prescribingAlgorithmic bias
Enhanced surveillanceLack of explainability
Accelerated antibiotic discoveryRisk of incorrect recommendations
Prediction of resistance mechanismsPrivacy and security concerns
Better infection controlHigh implementation costs
Support for underserved regionsRegulatory and legal challenges

What are the pros and cons of AI in environmental health?

Pros:

  • Improved Air Quality Monitoring and Prediction: AI can analyze data from monitoring systems, satellites, and other sensors to predict air pollution levels and identify pollution hotspots, improving public health planning.
  • Enhanced Disease Surveillance: AI can assess environmental data and predict disease outbreaks linked to climate, pollution, or vector-borne diseases leading to faster public health prevention and response strategies.
  • Better Risk Assessment: By processing large and complex datasets, AI can evaluate risks from environmental hazards such as toxic chemicals, contaminated water, or extreme health.
  • Climate Change and Health Modeling: Modeling relationships between climate variables and health outcomes, AI can predict heat-related illnesses, forecast wildfire smoke impacts, and assess climate-related disease risks.
  • Faster Environmental Data Analysis: AI relies on satellite imagery, air and water sensors, epidemiological databases, and genomic studies.
  • Improved Environmental Monitoring: Relying on sensors providing accurate, real-time data, AI can continuously monitor water and air quality, chemical contamination, and land-use changes.
  • Support for Public Health Decision-Making: AI can help identify emerging environmental threats and prioritize interventions based on predictive models.

Cons:

  • Dependence on Data Quality: AI systems are only as good as the data they use to train on and analyze.
  • Algorithmic Bias: Environmental sensors might be more abundant in affluent areas than in poor areas, leading to underestimation of risks in vulnerable populations, unequal allocation of resources, and environmental justice concerns.
  • Lack of Transparency: AI models operate as “black boxes” leading to reduced public trust, difficulty explaining predictions, and challenges for regulatory decision-making.
  • Risk of Incorrect Predictions: Rare events, rapidly changing environmental conditions, and insufficient training can lead to prediction failures.
  • Privacy Concerns: AI requires access to health records, mobility and sensor data, and geographic information. This raises concerns about data privacy, security, and ethical use of personal information.
  • High Implementation Costs: AI-based environmental monitoring systems require sensors and monitoring infrastructure, cloud computing resources, technical expertise, and ongoing maintenance. These costs can be substantial for public agencies and low-resource communities.
  • Regulatory and Ethical Challenges: AI-assisted decision-making requires transparency, appropriate government frameworks, and the ethical use of sensitive data.
  • Environmental Footprint of AI Itself: A notable paradox is that AI can help solve environmental problems while also consuming significant resources. High electricity use, water consumption for cooling systems, and increased demand for computing infrastructure are growing concerns about AI data centers’ energy, water, and land-use needs. Noise pollution is another concern.

Table 3: Summary of AI in Environmental Health

ProsCons
Better air and water quality monitoringDependent on high-quality data
Earlier disease and outbreak detectionAlgorithmic bias
Improved environmental risk assessmentLack of transparency
Faster analysis of large datasetsRisk of inaccurate predictions
Enhanced climate-health modelingPrivacy concerns
Real-time environmental surveillanceHigh implementation costs
Better public health decision supportRegulatory and ethical challenges
More efficient resource allocationAI data centers’ energy and water consumption and noise pollution.

Assessing Chat GPT’s Findings Regarding AI in One Health

After reviewing Chat GPT’s responses, a few themes stood out.

The first theme is the importance of data quality. The popular statement in computer science and data analytics, “garbage in, garbage out,” is as true with AI as it is with every other aspect of data analysis. Data quality depends on the people and systems collecting and managing it. For One Health purposes, this issue applies in clinical, research, veterinary, environmental, laboratory, radiology, pathology, agriculture, and other relevant settings.

Quality data must be accurate, valid, reliable, and representative to avoid bias and errors. This is particularly important for AI since it relies so much on large quantities of quality data to train and function properly. AI is being touted as providing faster and more accurate results, but it’s contingent on quality data. Critical issues such as AMR depend on accurate microbiologic and genomic data for proper AI analysis and predictions.

In the U.S., data in healthcare, public health, and environmental health is fragmented and of varying quality. Poor, rural areas have limited access to healthcare because of hospital closures, clinician shortages, and economic barriers. Although Chat GPT asserted that AI would solve this problem by increasing access to care, it acknowledged that implementation costs for healthcare systems would be prohibitive.

The U.S. doesn’t have universal health insurance or universal healthcare access, unlike other affluent countries. It ranks an appalling 62nd place in life expectancy for both sexes. It’s unclear if AI can improve this outcome without major changes in the insurance and access problems rooted in American healthcare.

In veterinary medicine, data quality will depend on the species that gets the most attention and care. In companion animal care, large corporate systems like Banfield Pet Hospital, are exploring AI’s potential for diagnostic and administrative purposes. AI is being adopted by large animal practitioners working in agriculture. Wearable devices provide livestock farmers data to monitor their welfare. AI is being explored to monitor wildlife for conservation purposes, but rapidly worsening biodiversity loss and climate change create challenges for systems trained on old data.

Interestingly, Chat GPT determined that AI in veterinary medical education was beneficial for test preparation, but didn’t mention the potential overreliance of technology that it highlighted as a concern for human medical education.

Public health and environmental health are chronically underfunded, and without support, will have difficulty generating quality data. States are cutting environmental health agencies. The Trump administration is dismantling a $368 million ocean monitoring system that provides critical environmental health information regarding the ocean, the ecosystem that impacts everyone on the planet. From a One Health perspective, this makes zero sense.

Privacy and security concerns are the second theme. The Health Insurance Portability and Accountability Act of 1996 (HIPAA) does not provide guidance when protected personal health information is shared with AI developers for training purposes to identify patterns and make predictions. Vast amounts of high-quality data require high volume which would make getting patient written authorization challenging. Even with the “de-identification” of patient data, privacy problems can occur.

This leads to the third theme of AI: legal and ethical challenges. Privacy and security concerns along with potential errors and bias in diagnosis and treatment recommendations raise the question of accountability. Who is responsible for errors? The AI developers, vendors, or clinicians? Addressing these concerns requires transparency. Legal expertise will have to be developed to address these issues.

The fourth theme is cost. AI development and implementation is expensive. Who will bear these costs? How would poor, rural hospitals pay for it? Would small group practices afford it? Will they be priced out?

The final theme is AI’s environmental impact, although Chat GPT never mentioned it as a negative until I specifically asked about environmental health. Currently, the US federal government is not conducting comprehensive environmental impact studies for AI data centers. Instead, the Trump administration has fast-tracked AI infrastructure and eased federal regulatory burdens. Some legislators, such as Representative Bonnie Watson Coleman (NJ-12) introduced a bill to study the environmental impact of AI on local communities.

State and local communities are fighting back against the mushrooming buildup of AI data centers because of concerns over their massive water and electricity consumption, noise pollution, and rising local utility bills. Grassroots activists have created a Stop Data Centers coalition to fight against their construction. Bipartisan opposition against them is building across the U.S.

No entity has determined how many AI data centers the country actually needs to maximize the benefits while minimizing the costs and risks.

My Findings from using Chat GPT as a Research Assistant

Perhaps it’s because I used the free version of Chat GPT that I found few of the references it provided to be useful. I conducted Google searches to find research articles that substantiated its claims. Confirming Chat GPT’s assertions consumed considerable time.

When I asked Chat GPT, “What are the pros and cons of AI in One Health?” the answers were essentially the same as when I had separated out the four questions involving human medicine, veterinary medicine, AMR, and environmental health. However, it didn’t mention AI’s deleterious environmental impact. (Please see Table 4). I have listed the references that Chat GPT provided for my One Health question below Table 4.

Other specific questions that I could have asked but didn’t because of time constraints included: The pros and cons of AI in food safety and security, sanitation and hygiene, vector-borne and zoonotic diseases, chronic diseases, mental health, agriculture, and climate change.

There’s no question that Chat GPT shortened my time researching the pros and cons of AI in One Health.

The benefits of AI in basic science research, drug discovery and development, and rapid diagnosis are clear. But AI comes with many costs and risks, particularly environmental risks. Growing concerns are being voiced at the grassroots and state levels.

From a One Health perspective, in an era of diminishing natural resources and worsening climate change, we must weigh the risks versus the benefits in the widespread adoption of AI.

Table 4: Summary of AI in One Health

ProsCons
Improved disease surveillanceData quality and interoperability issues
Early warning systemsLimited transparency (“black box” models)
Integration of diverse dataBias and inequity
Enhanced AMR monitoringPrivacy, ethical, and liability concerns and cybersecurity risks
Improved environmental monitoringInfrastructure requirements
Faster decision-making supportValidation challenges
Resource efficiency and optimizationDependence on cross-sector collaboration
Accelerated drug and vaccine discoveryRisk of misinformation and overreliance
Improved prediction of zoonotic disease emergenceRegulatory uncertainty

Chat GPT’s References and URLs