So there’s someone who has something medically wrong with them. The patient goes to a whole lot of doctors, but no one can figure it out. Until now, this may sound like an episode of House MD or one of those other medical dramas, but no, it’s real life. In the middle of the COVID-19 pandemic, a parent’s worst nightmare unfolded. A mother’s four-year-old son began exhibiting bewildering symptoms, with his health deteriorating.
She embarked on a journey that would span three arduous years and encompass over 17 doctor visits, but no answers were forthcoming. The kid’s symptoms were said to be as perplexing as they were distressing. He experienced pain, exhibited unusual chewing, and his growth seemed stunted. Yet, despite the parade of doctors and specialists, the root cause of his suffering remained enigmatic.
At this point, the mother turned to a seemingly unlikely last resort: OpenAI’s chatbot, GPT. This may have been a desperate attempt to uncover the elusive diagnosis for her son’s mysterious condition. The mother fed GPT the wealth of information shared during her son’s medical odyssey: MRI notes, new details of his symptoms, and even personal observations. And what did GPT do? It responded with a suggestion that would prove to be life-changing.
It proposed that the kid might be suffering from something called tethered cord syndrome. This is where the spinal cord gets abnormally attached to nearby tissues in the spine, causing potential health issues. So, this AI-generated diagnosis seemed to be a ray of hope. It seemed to resonate with the mother’s observations, especially the kid’s inability to sit cross-legged too long. This indicated a structural issue.
Buoyed by GPT’s insight, the mother reached out to a Facebook group dedicated to families grappling with the same condition, and she discovered stories mirroring her own son’s experience. On the power of GPT’s suggestion, she consulted a new neurosurgeon, and to her immense relief, the doctor confirmed the AI’s diagnosis. The specialists identified the exact location of the tethering on the child’s spinal cord.
What does this say about AI in healthcare? Why is GPT and other AI models valuable? It should not be considered infallible. Doctors and experts caution against complete reliance on AI for medical diagnosis. But the same could be said about WebMD. Maybe an AI like GPT could complement or be a co-pilot. Whether people marvel or criticize AI models like GPT, it is still considered impressive. Yet, these models can generate incorrect answers, a phenomenon that’s called a hallucination. So, AI could provide erroneous information, especially in complex medical cases.
Can it rival the diagnostic acumen of seasoned doctors? It was said that the latest iteration of GPT-4 answered US medical exam licensing questions accurately over 90 percent of the time. This could showcase its proficiency, especially in tackling standardized medical questions. AI models like GPT have the capacity to learn. A popular example of GPT’s prowess is citing the case of a lawyer who used GPT for work. GPT provided fake case laws to be presented.
Maybe the correct diagnosis of an elusive medical condition could be a counterweight to that narrative. That being said, while GPT’s contributions to medical diagnosis are promising, it is far from a substitute for human medical expertise. Collaborative healthcare might still be important, where doctors pull their knowledge and experiences. When AI models like GPT be a tool to augment medical expertise, not replace it? Will it be what a calculator was to accountants? Will it be an indispensable asset in the medical field? Would it aid in the timely diagnosis and treatment of conditions? What is the future of AI healthcare?
