"Gary Marcus has been warning the field for years that pure statistical approaches like LLMs probably wouldn't suffice for AGI. Half a trillion dollars later, it looks like maybe he had a point."

D'oh!