Shafaqna English- A Science Perspective says advanced AI models are getting closer to physician-level diagnostic reasoning, especially in text-based clinical cases. Studies cited in the article found GPT-4 reached near or exact diagnoses in up to 73% of cases, while OpenAI’s reasoning model o1-preview achieved 88.6% in clinicopathological cases and 67% accuracy in emergency triage scenarios, outperforming two expert physicians in some settings, according to News-Medical.
The article stresses that AI is not meant to replace doctors, but to support them in tasks such as clinical decision support, note generation, patient communication, administration, and research assistance. AI could reduce diagnostic errors, delays, and access barriers, but its real-world value is still uncertain.
Researchers argue that randomized trials, certification systems, and ongoing monitoring are needed before AI can be safely integrated more deeply into healthcare. They also warn that many evaluations remain text-only, failing to reflect the complexity of real clinical practice, and note risks including under-triage, bias, weak oversight, and lack of accountability.
Overall, the article concludes that AI is becoming highly capable in diagnosis, but accuracy alone is not enough — healthcare AI must also prove it is safe, fair, transparent, and clinically useful.
Source: News-Medical.net