Multimodal Large Models Toward AGI
摘要
The emergence of ChatGPT has transformed people’s perception of AI. For the first time, people witnessed an AI capable of performing various tasks requiring common sense or even professional knowledge, such as chatting, writing papers, and coding. Its impact rivals that of AlphaGo, gradually pushing the AI field toward AGI. Previously, AGI was considered an unattainable goal, with all prior AI breakthroughs falling under Narrow AI—where each AI could only perform a single task and lacked common sense. However, ChatGPT has changed this landscape, taking a significant step toward realizing AGI. Although current AGI remains relatively weak, its development is a topic worthy of in-depth exploration. Therefore, the author first attempts to analyze the challenges facing AGI and then proposes potential technical pathways, hoping to provide valuable insights for AGI research.