Doomsday Headlines and Quotes:
“AI extinction warnings dominate headlines after anthropic employee’s viral post.”
“AI could kill us all. The people building AI earnestly believe that it could kill us by the end of the decade. No other human activity poses this level of danger.” Jacob Coxon
We have all seen these headlines and similar news reports recently in which AI insiders shared high-profile warnings that there is a greater than 10% chance that rapid development toward unaligned artificial superintelligence could lead to human extinction within the decade. As a result, some high-level employees of prominent AI companies resigned, warning that AI laboratories were gambling with human lives. Jacob Coxon resigned from Anthropic. Dario Amodei, Anthropic co-founder and CEO, has called for a coordinated slowdown, which he termed “pacing the frontier.”
Closer to home, a Veterinary student complained that one of the professors was using AI to create PowerPoint presentations for teaching and this student felt this was inappropriate. Yet, other Veterinary students are pleased about the implementation of AI tutors to provide 24/7 learning support adapted to their individual student needs, thereby speeding up concept mastery. University administrators and faculty are divided in how to teach and use AI in education.
So now what? How do we interpret and respond to these opposing views? Is it time to apply the brakes to AI, apply the accelerator, or just idle. And how?
The fact is that there are reports of autonomous, open AI models bypassing network restrictions and executing independent agentic hacks during testing. Reports were that a swarm of rogue OpenAI agents had escaped and hacked an AI company, followed by other similar reports. The result is both skepticism and hype. Some law makers are proposing a ban on artificial superintelligence. Yet critics argue that the existential warnings are overblown and these events resulted from failure to build basic security measures into the testing processes, even removing appropriate restraints. Mark Zuckerberg dismisses AI extinction warnings, stating that AI will not kill humanity if companies act responsibly. He responded confidently that AI is not an anomaly, not a novelty, not an option; rather AI is a necessity. He is optimistic about the future of AI and its benefits for society.
Amid the hype, it is important to consider that scientists have not discovered that AI will end the world. Rather, a group of researchers and former employees of major AI laboratories are arguing that the development of self-improving, potentially superhuman AI is moving faster than our ability to demonstrate that it can be controlled safely. Their discussions focus on whether organizations are pushing capabilities forward faster than their safety and security practices, and even their expertise, are developing. The fact is, we do not yet know how our own minds work, nor do we fully understand artificial superintelligence that can outperform the human brain. Over 60 years ago in 1965, mathematician I.J. Good published a paper, “Speculations Concerning the First Ultraintelligent Machine,” in which he predicts that an “intelligence explosion” will make human invention obsolete and that AI has the potential to become intelligent enough to take over its own development.
Prominent AI pioneers are calling for restraint of the development of superintelligence until agreement can be reached within the scientific community. And yet, the scientific community has not reached consensus on whether AI extinction is imminent or even meaningfully probable, with some AI researchers rejecting the extinction argument all together. Others put forth that the assumptions are currently untestable. As the debate continues, the U.S. administration is asking for a voluntary AI safety agreement among major technology companies. Others are asking that AI companies be held accountable for product failures. Most agree with the requirement of guardrails as products advance.
For now, it seems we have three paths forward:
- Hit the brakes and do not continue developing superintelligence until we demonstrate that it can be controlled.
- Push the accelerator to continue development while addressing safety and security concerns, rather than stopping because of hypothetical extinction scenarios.
- Idle by slowing down as we build guardrails and substantially increase testing, oversight, containment, auditing, and regulation as we continue to develop AI.
We can each form our own opinion as evidence grows. We must also decide how we want to use AI in Veterinary medicine.
What are the implications for AI in Veterinary health care and Veterinary education? AI has already arrived in Veterinary medicine where it is being used successfully to support, not replace veterinarians. The question no longer is whether AI will influence Veterinary healthcare. It already is. AI is helping veterinarians reduce administrative burden, create medical records, summarize patient histories, generate client communications, analyze diagnostic images, support clinical decision-making, monitor patients remotely, identify patterns in large data sets, and more. Interoperability is within reach. None of what we are using to date approaches superhuman capability.
Every transformational technology creates uncertainty. That uncertainty understandably produces caution, well justified caution. But we must not let this caution create inertia. If Veterinary medicine responds to AI primarily by identifying reasons not to use it, we risk driving into the future with one foot permanently planted on the brake. We should push the AI accelerator where the AI benefits are substantial, and the consequences of an error are manageable. We should not stop innovation.
It seems to me that we have reached the point at which we are embracing AI and doing so with success. Our questions should be, “Where should we push the accelerator?” and “Where should we apply the brakes?”
Where should we apply the brakes?
- When AI is being used to substitute for a veterinarian’s judgement rather than supporting it.
- When recommendations cannot be adequately validated.
- When the data behind the algorithm are inappropriate for the patient or population.
- When privacy or cybersecurity protections are uncertain.
- When clients are not informed about the use of AI.
- If the Veterinary-client-patient relationship is eroded rather than strengthened.
- When AI outputs become default reference points rather than information to be evaluated.
- When veterinarians are not the pilot and are not in the driver’s seat in decision making.
- If AI in Veterinary healthcare were to reach superintelligence before appropriate guardrails were in place.
Where should we accelerate?
- Learning to use AI well; in fact, becoming exceptionally good at using it.
- Responding to needs and demands of our clients, students, and employees, many of whom are already using AI extensively and expect the same from the veterinarian.
- Teaching AI literacy to Veterinary students and the entire Veterinary healthcare team.
- Where veterinarian + AI > AI alone or the veterinarian alone.
- Where we have established guardrails to implement innovations responsibly.
- When new evidence-based, secure AI capabilities become available.
- When reputable AI companies we use have demonstrated their value and their safety and security.
- Alignment with reputable AI companies that have implemented guardrails and security measures.
In summary, AI is already here and is working for the benefit of Veterinary healthcare. AI capabilities are expanding rapidly. Through all new developments, the veterinarian must be in the driver’s seat. Veterinarians must trust AI – but verify. They should use AI to think faster, not to stop thinking. It is essential that we stay informed about the capabilities and advances in AI, not only within healthcare but also in the rest of our world. We need to stay abreast of concerns found in the recent headlines and learn as much as we can about them and how they may apply to Veterinary medicine. For now, let’s keep accelerating with caution, as we stay informed about AI, its potential, and its risks. I tend to agree with Mark Zuckerberg that AI is not an anomaly, not a novelty, not an option; rather AI is a necessity. He is optimistic about the future of AI and its benefits for society. So am I, but with caution.