OpenAI’s recent GPT-5 launch, initially seen as a breakthrough in AI, quickly turned into a crisis as users expressed dissatisfaction with the new model routing system. Many felt a lack of control, leading to accusations that OpenAI was not transparent. The routing setup integrates several model variants, prompting users to experience inconsistent responses from different models addressing the same query. While this system offers potential benefits, including resource efficiency by reusing older models, it has revealed significant challenges during GPT-5’s rollout. OpenAI’s swift reinstatement of GPT-4o for users reflects a desire to rebuild trust. Experts suggest model routing may be the future of AI despite its shortcomings, as it addresses physical limitations in training larger models. The hyped promise of artificial general intelligence (AGI) remains elusive, with AI scholars suggesting a hybrid approach may be necessary, balancing specialized models without pursuing a singular all-powerful system.
Source link
