Back to today’s briefing

Story perspectives

Apple Researchers Question AI Models' Reasoning Abilities

6/15/2025

49 6

Summary

Story summary

Apple researchers have raised alarms about leading AI models, such as OpenAI's o3 and Anthropic's Claude 3.7, revealing their struggles with fundamental reasoning tasks like the Tower of Hanoi. With accuracy sinking below 80% for seven disks and further declining for eight, these findings cast doubt on the advancements toward achieving true artificial general intelligence (AGI).

1 of 1

Story summary

Apple researchers have raised alarms about leading AI models, such as OpenAI's o3 and Anthropic's Claude 3.7, revealing their struggles with fundamental reasoning tasks like the Tower of Hanoi. With accuracy sinking below 80% for seven disks and further declining for eight, these findings cast doubt on the advancements toward achieving true artificial general intelligence (AGI).