AI Limitations
Use abstention and human review when evidence is inadequate.
How it works
A system’s observed task performance does not establish consciousness, feelings or a theory of mind. A robot also needs sensing, control and physical safety engineering beyond a model score. Claims about job replacement require careful evidence and cannot be inferred from a toy example. This small policy simply routes uncertain scores for review; it is not an autonomous deployment.
Run the code
Try in playground loads this complete standard-library Python example without executing it. Choose Run when ready. Local review uses Python 3.12.14; the existing browser editor uses Pyodide 0.27.7. These small fictional examples need no account, external data, model-provider key or database service.
Example
import math
def route(score):
if score is None or isinstance(score,bool) or not isinstance(score,(int,float)) or not math.isfinite(score) or not 0 <= score <= 1:
return "review"
return "suggestion" if score >= 0.85 else "review"
print([route(score) for score in [0.92,0.62,None]])
Expected result
['suggestion', 'review', 'review']
A suggestion is still subject to the application’s review policy. The threshold is illustrative and the score is not shown to be calibrated. Human oversight also needs time, information and genuine authority to refuse a recommendation.