The Truth No One Wants: Artificial Intelligence Fails in 80% of Early Diagnoses… Human Medicine Still Prevails
April 14, 20261027 ViewsRead Time: 2 Minutes
Font Size
16
A scientific study published in JAMA Network Open showed that artificial intelligence robots failed in over 80% of medical cases during the early stages of diagnosis, when information is incomplete or symptoms are unclear.
These results highlight a real gap between the high expectations for this technology and the reality of its performance in the most sensitive medical moments.
Why Does AI Fail in the Early Stages of Disease?
The problem lies in that early diagnosis relies on analyzing incomplete and variable data, which confuses intelligent models that tend to:
quickly narrow down possibilities
choose one diagnosis instead of building a list of possibilities
rely on complete data to make accurate decisions
This contrasts with the methodology of doctors, which is based on gradual assessment and cumulative analysis.
When Data is Complete… AI Accuracy Increases Remarkably
In contrast, the study showed a significant improvement in performance when complete information was available, with the error rate dropping to less than 40%, while the accuracy of some models exceeded 90%.
This means that these systems are effective in:
supporting final diagnoses
analyzing test results
confirming clear cases
But they are not qualified as a starting point for the diagnostic process.
A global test included major AI companies
The study included 21 models from major companies, including:
OpenAI
Google
Anthropic
xAI
DeepSeek
They were tested across 29 hypothetical clinical cases presented progressively to accurately simulate medical reality.
Warnings Against Relying on Technology Without Medical Supervision
Researchers warned that relying solely on artificial intelligence could lead to serious errors, especially in cases where symptoms are ambiguous or incomplete.
The lead researcher emphasized that these models “shine when the picture is complete, but stumble when the story begins.”
Technology Companies: Assistive Tools, Not a Replacement for Doctors
For their part, the developing companies stressed that these systems are not designed to independently diagnose diseases:
Anthropic directs users to specialists
Google integrates alerts to verify information
OpenAI prohibits the use of its services as a substitute for medical consultation
The Hard Truth… A Doctor Cannot Be Replaced
Despite all the technological advancements, this study proves that artificial intelligence is still far from playing the role of a doctor, especially in early stages that require precise human expertise.
The truth no one wants: technology helps… but it does not diagnose alone.
