Select Language:
Renowned researcher Yang Likun has recently voiced strong criticism of the current trajectory in the development of large language models (LLMs). According to him, the prevailing approach may be fundamentally flawed. Instead, Yang advocates for a shift toward the JEPA world model, which he believes holds the key to achieving Artificial General Intelligence (AGI).
In a recent discussion, Yang argued that the emphasis on scaling up language models alone might be misguided. He contends that despite their impressive capabilities, these models lack a comprehensive understanding of the world, limiting their potential for true AI generality. Instead, he champions the JEPA framework — a structured world model designed to mimic human reasoning and perception more accurately.
Yang emphasizes that the journey toward genuine AGI requires more than just bigger datasets and more parameters. It demands a paradigm that integrates a deeper, more interconnected understanding of the environment. The JEPA model, which centers on creating a dynamic, holistic world representation, reflects this philosophy. By doing so, it aims to bridge the gap between narrow AI and the kind of flexible intelligence that humans exhibit.
His critique resonates with a growing number of experts who believe that the current focus on scaling language models may overlook essential aspects of cognition. The debate underscores the need for innovative approaches in AI research—approaches that go beyond mere size and statistical correlations.
While the future of AI remains uncertain, Yang Likun’s perspective injects fresh thinking into ongoing discussions. As the community aims for the elusive goal of AGI, the emphasis on building unified world models like JEPA could represent a promising path forward.
ChatGPT
Perplexity
Gemini AI
Grok AI
