Dr Amy Perkins
AI in Dermatology
I have a special interest in how artificial intelligence can support dermatology and wider healthcare, particularly in remote and rural settings. My work includes contributing to HealthBench, a project developing high-quality benchmarks to evaluate the safety and accuracy of AI tools in healthcare.
I collaborate with clinicians, researchers, and developers to explore how AI can assist with diagnosis, triage, and patient education, while ensuring safety, ethics, and equity remain central. I am also involved in discussions around regulation, transparency, and public trust in AI systems, with a focus on how technology can complement — not replace — human expertise.
Through speaking engagements, panel discussions, and writing, I aim to bridge the gap between clinicians and technologists, helping shape AI solutions that are clinically relevant and evidence-based.
My Work with OpenAI
HealthBench is an open-source project created by OpenAI to set new standards for testing artificial intelligence in healthcare. Instead of relying on exam-style questions, it uses 5,000 real-world conversations that mirror the kinds of discussions patients and clinicians have every day. This makes it one of the most realistic and useful ways to see how AI might perform in genuine clinical settings.
What makes HealthBench unique is that the evaluation criteria were designed by over 260 doctors from around the world, including myself. Together, we developed thousands of measures that look not only at whether the information is correct, but also whether it is safe, empathetic, and genuinely useful in practice.
The project is built on three guiding principles: tests should be meaningful for patient care, trustworthy in how they reflect medical judgement, and open enough that AI systems can continue to improve. Early results are promising, but they also highlight how far there is to go before AI can be trusted to act without human oversight.
HealthBench has already been recognised as a turning point for healthcare AI, helping move the field towards safer, more ethical, and clinically relevant tools for the future.
