\u672c\u65e5\u306e\u6ce8\u76eeAI\u30fb\u30c6\u30c3\u30af\u30cb\u30e5\u30fc\u30b9\u3092\u3001\u5c02\u9580\u7684\u306a\u5206\u6790\u3068\u5171\u306b\u304a\u5c4a\u3051\u3057\u307e\u3059\u3002<\/p>\n
\u3053\u306e\u8a18\u4e8b\u306fAI\u306b\u3088\u3063\u3066\u81ea\u52d5\u751f\u6210\u30fb\u5206\u6790\u3055\u308c\u305f\u3082\u306e\u3067\u3059\u3002AI\u306e\u6027\u8cea\u4e0a\u3001\u4e8b\u5b9f\u8aa4\u8a8d\u304c\u542b\u307e\u308c\u308b\u53ef\u80fd\u6027\u304c\u3042\u308b\u305f\u3081\u3001\u91cd\u8981\u306a\u5224\u65ad\u3092\u4e0b\u3059\u969b\u306f\u5fc5\u305a\u30ea\u30f3\u30af\u5148\u306e\u4e00\u6b21\u30bd\u30fc\u30b9\u3092\u3054\u78ba\u8a8d\u304f\u3060\u3055\u3044\u3002<\/p>\n<\/div>\n<\/div>\n
\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb\uff08LLM\uff09\u306e\u9032\u6b69\u306f\u3001\u4e88\u6e2c\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u3092\u57fa\u76e4\u3068\u3059\u308b\u30c8\u30e9\u30f3\u30b9\u30d5\u30a9\u30fc\u30de\u30fc\u30e2\u30c7\u30eb\u306e\u6700\u9069\u5316\u306b\u57fa\u3065\u3044\u3066\u3044\u308b\u304c\u3001\u73fe\u5728\u306e\u30e2\u30c7\u30eb\u306f\u884c\u52d5\u3001\u968e\u5c64\u7684\u69cb\u9020\u3001\u30a8\u30d4\u30bd\u30fc\u30c9\u8a18\u61b6\u3068\u3044\u3063\u305f\u795e\u7d4c\u79d1\u5b66\u3067\u91cd\u8981\u8996\u3055\u308c\u308b\u8981\u7d20\u3092\u7121\u8996\u3057\u3066\u3044\u308b\u3002\u672c\u8ad6\u6587\u3067\u306f\u3001\u5b89\u5168\u3067\u89e3\u91c8\u53ef\u80fd\u3001\u30a8\u30cd\u30eb\u30ae\u30fc\u52b9\u7387\u304c\u9ad8\u304f\u3001\u4eba\u9593\u306e\u3088\u3046\u306aAI\u3092\u5b9f\u73fe\u3059\u308b\u305f\u3081\u306b\u3001\u3053\u308c\u3089\u306e\u8981\u7d20\u3092\u884c\u52d5\u3001\u968e\u5c64\u7684\u69cb\u9020\u3001\u30a8\u30d4\u30bd\u30fc\u30c9\u8a18\u61b6\u3092\u7d71\u5408\u3057\u305f\u57fa\u76e4\u30e2\u30c7\u30eb\u306b\u7d44\u307f\u8fbc\u3080\u3053\u3068\u3092\u63d0\u6848\u3059\u308b\u3002\u3053\u308c\u306b\u3088\u308a\u3001\u5e7b\u899a\u3001\u7406\u89e3\u306e\u6b20\u5982\u3001\u30a8\u30fc\u30b8\u30a7\u30f3\u30b7\u30fc\u306e\u6b20\u5982\u3001\u89e3\u91c8\u53ef\u80fd\u6027\u306e\u4f4e\u3055\u3001\u30a8\u30cd\u30eb\u30ae\u30fc\u52b9\u7387\u306e\u60aa\u3055\u3068\u3044\u3063\u305f\u73fe\u5728\u306eLLM\u306e\u6b20\u70b9\u3092\u514b\u670d\u3067\u304d\u308b\u53ef\u80fd\u6027\u304c\u3042\u308b\u3002<\/p>\n
\ud83d\udc49 arXiv \u3067\u8a18\u4e8b\u5168\u6587\u3092\u8aad\u3080<\/a><\/strong><\/p>\n While advances in Large Language Models (LLMs) are based on optimizing transformer models using predictive coding, current models overlook crucial elements from neuroscience: actions, compositional structure, and episodic memory. This paper proposes integrating these components into foundation models to achieve safe, interpretable, energy-efficient, and human-like AI. By incorporating actions at multiple scales, a compositional generative architecture, and episodic memory, the authors argue that current LLM deficiencies such as hallucinations, lack of grounding, missing sense of agency, poor interpretability, and energy inefficiency can be addressed.<\/p>\n<\/blockquote>\n<\/div>\n \u30b7\u30fc\u30b1\u30f3\u30b9\u69cb\u9020\u306f\u3001\u8a00\u8a9e\u3001\u904b\u52d5\u3001\u610f\u601d\u6c7a\u5b9a\u306a\u3069\u3001\u4eba\u9593\u306e\u8a8d\u77e5\u3084\u884c\u52d5\u306b\u304a\u3051\u308b\u91cd\u8981\u306a\u7279\u5fb4\u3067\u3042\u308a\u3001\u4eba\u5de5\u77e5\u80fd\u3078\u306e\u5fdc\u7528\u306b\u304a\u3044\u3066\u3082\u4e2d\u5fc3\u7684\u306a\u6027\u8cea\u3092\u6301\u3064\u3002\u672c\u8ad6\u6587\u3067\u306f\u3001\u30c9\u30e1\u30a4\u30f3\u306b\u4f9d\u5b58\u3057\u306a\u3044\u5f62\u3067\u30b7\u30fc\u30b1\u30f3\u30b9\u5b66\u7fd2\u3068\u51e6\u7406\u3092\u8a55\u4fa1\u3057\u3001\u8a08\u7b97\u7406\u8ad6\u3068\u306e\u9023\u643a\u3092\u63d0\u4f9b\u3059\u308b\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3068\u3057\u3066\u3001SymSeq\uff08\u69cb\u9020\u5316\u30b7\u30f3\u30dc\u30ea\u30c3\u30af\u30b7\u30fc\u30b1\u30f3\u30b9\u306e\u751f\u6210\u3068\u5206\u6790\u30c4\u30fc\u30eb\uff09\u3068SeqBench\uff08\u30eb\u30fc\u30eb\u30d9\u30fc\u30b9\u306e\u30b7\u30fc\u30b1\u30f3\u30b9\u51e6\u7406\u30bf\u30b9\u30af\u306e\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u30b9\u30a4\u30fc\u30c8\uff09\u3092\u63d0\u6848\u3059\u308b\u3002SymSeqBench\u306f\u3001\u5f62\u5f0f\u8a00\u8a9e\u7406\u8ad6\uff08FLT\uff09\u306b\u57fa\u3065\u304d\u3001\u8a8d\u77e5\u79d1\u5b66\u3001\u5fc3\u7406\u8a00\u8a9e\u5b66\u3001\u30cb\u30e5\u30fc\u30ed\u30e2\u30eb\u30d5\u30a3\u30c3\u30af\u30b3\u30f3\u30d4\u30e5\u30fc\u30c6\u30a3\u30f3\u30b0\u3001AI\u306a\u3069\u306e\u5206\u91ce\u3067FLT\u306e\u6982\u5ff5\u3092\u9069\u7528\u3057\u3001\u5b9f\u9a13\u3092\u6a19\u6e96\u5316\u3059\u308b\u305f\u3081\u306e\u4fbf\u5229\u3067\u5b9f\u7528\u7684\u306a\u65b9\u6cd5\u3092\u63d0\u4f9b\u3059\u308b\u3002<\/p>\n\n
\n
SymSeqBench: a unified framework for the generation and analysis of rule-based symbolic sequences and datasets<\/h2>\n
\n
\u5c02\u9580\u30a2\u30ca\u30ea\u30b9\u30c8\u306e\u5206\u6790<\/h3>\n