According to reports, the head of Meta AI stated that current artificial intelligence models lack key characteristics of human intelligence and claimed that the company's latest model will address this issue.
Business Insider reported on May 26 that at the AI Action Summit held in Paris earlier this year, Meta's Chief AI Scientist Yann LeCun stated, "Intelligent behavior has four fundamental characteristics that every animal, or relatively intelligent animal, can achieve, and humans can certainly do so."
"Understanding the physical world, having persistent memory, being able to reason, and being able to plan complex actions, especially hierarchical planning," LeCun said.
He noted that the large language models (LLMs) driving popular AI chatbots have not yet reached this threshold, and "integrating these capabilities will require changing the way they are trained."
LeCun mentioned that some of the largest AI and tech giants are "patching together" capabilities onto existing models in the race for AI dominance.
Meta has been experimenting with a system called Retrieval-Augmented Generation (RAG), which is a method of enhancing LLM output using external knowledge sources.
In February, it released V-JEPA, a non-generative model that learns by predicting missing or obscured parts of videos.
LeCun believes that "world-based models" will be a better approach, as these models will be trained on real-world scenarios and possess higher cognitive abilities than current pattern-based AIs.
This concept involves models that can imagine taking actions and predict the resulting state of the world. Given the infinite unpredictability of the world, LeCun believes that training must occur through abstraction, reflecting how humans understand the physical world.
Meanwhile, according to Insider on May 26, Meta's AI research team is experiencing significant talent loss, particularly among researchers who created the original Llama model in 2023.
Only three of the original 14 Llama authors remain at Meta, with many having joined Mistral, a Paris-based startup co-founded by former Meta researchers and key Llama architects.
The report added that Meta's recently released Llama 4 has received a lukewarm response from developers, many of whom are now turning to faster-moving competitors that have specialized reasoning models, such as OpenAI's GPT-4o, Google's Gemini 2.5 Pro, and Anthropic's recently launched Claude 4 Sonnet.
On May 15, The Wall Street Journal reported that Meta is delaying the launch of its flagship AI LLM Llama 4 "behemoth."
Related: Researchers say ChatGPT model refused shutdown requests during testing
Original: “Report: Meta AI Head Claims Current AI Lacks ‘Intelligent Behavior’”
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
