Just wondering out loud…can bias ever really be “controlled”if its source is the subjective human user? And if user interpretations, expectations and contexts shift constantly, how could a system reliably recognize the “correct”pattern at an early stage?
1 Like
Counter-question - Does it therefore make sense to work with ambiguities and vagueness?
At this point, I would like to refer once again to my posts in this thread and thus to my core argument, which I have already explained sufficiently.
To this question: sure!
If AI systems are already capable of doing such things TODAY! Than it’s no since fiction anymore.
-
AI systems now match or outperform experienced dermatologists in image-based skin cancer detection, especially for melanoma.
-
Sensitivity (true-positive rate) is often higher for AI (≈ 85–95 %), meaning fewer missed melanomas compared to clinicians.
-
Specificity can be slightly lower for AI, leading to more false positives than expert dermatologists in some studies.
-
Large meta-analyses and multicenter studies show AI performance is comparable or superior to dermatologists in overall diagnostic accuracy.
-
Most evidence comes from retrospective image datasets; real-world clinical validation is improving but still ongoing.
-
Best results are achieved when AI is used as a decision-support tool, not as a standalone diagnostic system.
-
Clinically, AI has strong potential to increase early detection and patient safety, especially in screening settings.
1 Like
Hi,
Your concern is thoughtful and speaks to pressures on framework and perspective formation.
The problem (and I do not mean to be ornery) .
There is a commercially available camera that can capture 20 million frames a second. For entities that can process visuals that fast, humans are almost frozen. What happens to perspective, consent of the governed, human framework formation or attitudes then? (for unenhanced humans)