Apple researchers: AI models are still far from reasoning at the AGI level.

The race to develop general artificial intelligence (AGI) still has a long way to go, according to Apple researchers who argue that leading AI models such as ChatGPT and Claude still struggle with arguments. Although there have been (LRMs) large models of argumentation, their capabilities and limitations are still poorly understood. Current assessments mainly focus on the accuracy of the final answer without providing a look at the ability to reason.

Research shows that LRM does not generalize argumentation effectively and often experiences a collapse in accuracy as complexity increases. Researchers conclude that these models merely mimic reasoning without actually capturing it, which does not meet the requirements of AGI. AGI remains a distant goal, although some industry leaders believe it could be achieved in the coming years.