Concentric AI CTO says AI still needs human judgment and oversight
Overview
Madhu Shashanka, CTO of Concentric AI, says AI remains "jagged": it performs well on tasks that can be checked automatically but fails where no automated test can judge the right answer.
He argues that the work that remains for people is judgment-based, including testing AI outputs, setting limits on AI agents, and supervising them.
Shashanka also warns that automating entry-level routine jobs removes the work where people learn this judgment, and says that loss undermines the skills needed for oversight.
Written by AI from the articles below · updated Oct 20, 8:44 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- The Guardian · AINewsAI still needs human judgment as tasks shift from automation to oversight
AIMadhu Shashanka, CTO of Concentric AI, argues that AI remains "jagged," excelling at checkable tasks but failing where no automated test can judge the right answer. He says the emerging work is judgment-based: testing outputs, setting limits on agents, and supervising them. He warns that automating entry-level routine jobs destroys the training ground where that judgment is built.
Heat Trend
Not enough continuous observations to show a trend yet.