以ä¸ã¯ä¸è¨ã®2ã¤ã®è«æï¼ãPICSY: Propagational Investment Currency SYstemãã¨ãBirdwatchãï¼ãã¤ãªãé åã®è¦ç¹ã¨ã¾ã¨ãã§ããã
以ä¸ã¯ä¸è¨ã®2ã¤ã®è«æï¼ãPICSY: Propagational Investment Currency SYstemãã¨ãBirdwatchãï¼ãã¤ãªãé åã®è¦ç¹ã¨ã¾ã¨ãã§ããã
ç±³OpenAIã¨AIå®å ¨æ§ã®ç 究çµç¹ã§ããç±³Apollo Researchã¯12æ5æ¥ï¼ç¾å°æéï¼ãOpenAIãåæ¥å ¬å¼çããªãªã¼ã¹ããLLMã®ão1ãã«ã¤ãã¦ã®èå¯ãçºè¡¨ãããããããão1ã«é¢ãã¦å¹¾ã¤ãã®æ¸å¿µäºé ãææãã¦ããã o1ã¯ãOpenAIã9æã«çºè¡¨ããLLMãåçãçæããåã«æèã®é£éãçã¿åºããã¨ã§ãè¤éãªæ¨è«ã¿ã¹ã¯ãå®è¡ã§ããã¨ãããã®ã ããã¾ãã¾ãªæ¦ç¥ã試ããããééããèªèããããæèããã»ã¹ãæ´ç·´ããããããã¨ã§ãå¾æ¥ã®ã¢ãã«ã¨æ¯è¼ãã¦ãããæç¨ãªåçãæä¾ã§ããããã«ãªã£ãã¨OpenAIã¯èª¬æãã¦ããã ã ãApollo Researchã¯ão1ããschemingãï¼çç¥ãéçºè ãã¦ã¼ã¶ã¼ã®æå³ã¨ç°ãªãç®çãå¯ãã«è¿½æ±ããè½åï¼ã«ã¤ãã¦è©ä¾¡ããã¨ããão1ã以ä¸ã®ãããªè¡åã示ããã¨ãæããã«ãªã£ãã¨ãã¦ããã éçºè ãã¦ã¼ã¶ã¼ã§ã¯ãªãão1èªèº«ã®ç®æ¨ã«
ãã®è¨äºã¯Haskell Advent Calendar 2024ã®6æ¥ç®ã®è¨äºã§ãã æ¦è¦ Haskellåãã®LLMã®ãã¬ã¼ã ã¯ã¼ã¯ (Intelli Monad)ãä½æããèªç«ã¨ã¼ã¸ã§ã³ãã§ããã¢ããªã±ã¼ã·ã§ã³ã«çµã¿è¾¼ã¿ã¾ããã ãã®ãã¬ã¼ã ã¯ã¼ã¯ã¯åå®å ¨ãªFunction CallingãStructured Outputãæä¾ãããã£ããã®å±¥æ´ãã¢ããã§ç®¡çããå±¥æ´ã®æ°¸ç¶åãè¡ãã¾ããLLMã«ã¯OpenAIãå©ç¨ããå±¥æ´ã®æ°¸ç¶åã«ã¯Persistentã©ã¤ãã©ãªãå©ç¨ãã¦ãã¾ããLLMã®å ¥åºåãåå®å ¨ã§ãããããHspecãªã©ã®æ¢åã®ã¦ããããã¹ãã§ãã¹ããå¯è½ã§ãã ç®ç Haskellã使ã£ã¦LLMï¼å¤§è¦æ¨¡è¨èªã¢ãã«ï¼ãèªç«ã¨ã¼ã¸ã§ã³ãã§ããã¢ããªã±ã¼ã·ã§ã³ã«çµã¿è¾¼ããã¨ãç®çã§ãã çµã¿è¾¼ã¿å¯¾è±¡ã®ã¢ããªã±ã¼ã·ã§ã³ã«ã¯ã次ã®äºã¤ã®ãã¿ã¼ã³ãåå¨ãã¾ããä¸ã¤ã¯ãã¥ã¼ãã³ã»ã¤ã³ã»
making-the-most-of-local-llms.ipynb Loading Sorry, something went wrong. Reload? Sorry, we cannot display this file. Sorry, this file is invalid so it cannot be displayed.
ãã¡ãã¯ãLLMã»LLMæ´»ç¨ Advent Calendar 2024ãã®ï¼æ¥ç®ã®è¨äºã§ã! 対象èªè LLMããªãã¨ãªã触ãå§ãã人 ã¯ããã« LLMã¯ãããã次ã«ããåèªãäºæ¸¬ãã¦åºåãããã¨èª¬æããã¾ããããã¯å ¥åããå 容ãå ã«ãåºç¾ç¢ºçãäºæ¸¬ããã¢ãã«ã§ããããããã®ä¸ããå¯è½æ§ã®é«ãåèªãé¸æãããã¨ã§æç« ã®çæãã§ããããã§ãããã ããã¶ã£ã¡ããã¦è¨ãã¨APIã¬ãã«ã§å©ãããllama.cppãéã¶åã«ã¯ããã¯ã©ããããã¦ãããåºæ¬çã«ã¯ãã¾ãèããå¿ è¦ã¯ãªãããã«ä½ããã¦ããã¨æãã¾ãããå人çã«ãLLMãç¨ãããããã¯ããä½ãéã¯ããã¯å¼·ãæèããå¿ è¦ã¯ãªãã¨ãèãã¦ãã¾ãã ä¸æ¹ã§ããããå°ãææ¡ãã¤ã¤ç¢ºçåå¸ãè¦ãããããå¼ã£ãããããã¨ã§LLMãç¨ãã表ç¾ã®å¯è½æ§ãæ¢ããã¨ãã§ããã¯ãã§ãããã®è¨äºã¯LLMã®åºåå 容ããã³åãã¼ã¯ã³ãã¨ã®ç¢ºçãçºãã¤ã¤ãä½ã«ä½¿ãããã
LLMããã£ã¦ãããªãã§å®æãã¼ã¹ã§å¦ãã ãããã®è¦ç´ ã2åã3åãããããã¨ããã®æ¡ä»¶ã¯ãã°ããããããªããã¨æãããªã¹ãã§ããèªåã®åå¿ãããã¦æ¸ãæ®ãã¦ããã¾ããã¾ããä»å¾ãè²ã¦ã¦ããããã§ãããã¨ä½ãããããªï¼ 1.ãã¼ã¯ã³æ°ãè¶ ããå ¥åãå¿ è¦ã¨ããã¿ã¹ã¯ âã¤ã³ããããã¼ã¿ããã¯ãã«DBåããªãã¨ãããªããã精度ãä¸ãã 2.è¤æ°ãã¡ã¤ã«ã®ã¤ã³ããããå¿ è¦ã¨ããã¿ã¹ã¯ âãã¡ã¤ã«éã®é¢ä¿ãæ£ç¢ºã«ææ¡ã§ããªãããããã«ã·ãã¼ã·ã§ã³çºççãä¸ãã 3.æ¥åãå·¥ç¨ã横æããã¿ã¹ã¯ âç°ãªãæèãæ£ç¢ºã«ææ¡ã§ããªãããããã«ã·ãã¼ã·ã§ã³çºççãä¸ãã 4.ã¤ã³ãããã®å½¢å¼ï¼æ§å¼ãè¤éãªã¿ã¹ã¯ âç»åãå³ãªã©ã¯ç²¾åº¦ãä¸ãã(ã¤ã³ããªã¸ã§ã³ããªOCRããã«ãã¢ã¼ãã«LLMã§ãããã£ããå³ããæ¦ãããã) 5.ããã¸ã§ã¯ãã§ä½¿ããLLMã®ã¢ãã«ãã¬ã¬ã·ã¼ã¢ãã«(GPT4以ä¸)ã§ãã âã¢ãã«èª
We believe that studying these agents on a large scale offers valuable insights into their behaviors, capabilities, and potential risks. Research with us Rigorous research takes time and resources. With advanced GPU access and a dedicated team, we explore the frontier research topics by balancing urgency and patience. Join our ongoing projects or test new ideas with us, reach out via email for mor
ã¯ããã« ç¾å¨ã社å ã§ã®æ¥åã®ä¸ç°ã¨ãã¦ãæ¥åå¹çåã®ããã®å ¨ç¤¾çãªçæAIã®æ¨é²æ½çã主å°ãã¦ãã¾ããæ½çãå§ããã®ãä»å¹´ã®2æé ãªã®ã§ã11æç¾å¨ã§ç´10ã¶æã¨ããã¨ããã§ããçµæã¨ãã¦ãã¢ã³ã±ã¼ãåçè ç´87åã®å 6å²ãé±1å以ä¸ã¯çæAIãæ´»ç¨ãã¦ããã¨ããç¶æ ã¾ã§ãã£ã¦ããã¾ããã ãã®è¨äºã§ã¯åãçµã¿ã®ãã£ããã¨ãå®éã®æ½çã¨ãã®çµæã«ã¤ãã¦ã¢ã³ã±ã¼ãçµæãªã©ã交ãã¦èª¬æã§ããã°ã¨æãã¾ãã èæ¯ ããããçæAI(ããã§ã¯ç»åçæAIã¯ãã£ããé¤ãã¦ããã¾ãï¼ã¯ITç³»ä¼æ¥ã®ã¿ãªãããã2å¹´ã§éå浸éãã¦ãã¾ãããæåã©ãããæãã㨠ChatGPT (OpenAI) Claude (Anthropic) Gemini (Google) GitHub Copilot (Microsoft) ãããã§ãããããå ãã¦ãæè¿ã¯ãªã¼ãã³ãªçæAIâ¦â¦ã¨ãããã大è¦æ¨¡è¨èªã¢ãã«ï¼LLM
æ¬è¨äºã§ã¯ãRAGã®æ§è½ãé«ããããã®ãPlanÃRAGãã¨ããææ³ã«ã¤ãã¦ããã£ããç解ãã¾ããæ ªå¼ä¼ç¤¾ãã¬ãã¸ã»ã³ã¹ã¯ãã¨ã³ã¿ã¼ãã©ã¤ãºä¼æ¥åãã«RAGãæä¾ãã¦ããã¹ã¿ã¼ãã¢ããã§ãã ãã®è¨äºã¯ä½ ãã®è¨äºã¯ãRAGã®æèæ¶ããåé¡ãå æããæ°ææ³ãPlanÃRAGãã®è«æ[1]ã«ã¤ãã¦ãæ¥æ¬èªã§ç°¡åã«ã¾ã¨ãããã®ã§ãã ä»åããããããRAGã¨ã¯ï¼ãã«ã¤ãã¦ã¯ãç¥ã£ã¦ããåæã§é²ã¿ã¾ãã確èªããå ´åã¯ãã¡ãã®è¨äºããåèä¸ããã æ¬é¡ ãã£ãããµããªã¼ PlanÃRAGã¯ãRAGã®ç²¾åº¦ãä¸ããããã®æ°ããææ³ã§ããã¢ã¼ã«ã大å¦ã¨Microsoft Researchã®ç 究è ãã«ãã£ã¦2024å¹´10æã«ææ¡ããã¾ããã ãã£ããè¨ãã¨ãPlanÃRAGã¨ã¯ããè¨ç»ãç«ã¦ã¦ããæ¤ç´¢ãããææ³ã§ããPlanÃRAGã§ã¯ãã¦ã¼ã¶ã¼ã®è³ªåããã¾ãæåã«å°ããªåä½ã«å解ããã®å¾ããããã®é¢ä¿æ§
ã¯ããã« 10æ12æ¥çºå£²ã®ãSoftware Designç·éç·¨ã2018ï½2023ãã«ãããç¹éãçæAIæ代ãçãæãITã¨ã³ã¸ãã¢ã®çããæè¡ãã§ã第1ç« ãããã°ã©ãã³ã°ããæ å½ããã¦ããã ãã¾ããã éèªçºå£²ããä¸ã¶æè¿ãçµã£ã¦ããã®ã§ããææãé¸ããæãããã®ã§ãããæå±ä¼ç¤¾ã®ååãåºãã¦æ¸ããè¨äºã§ãããã®ã§ããã®å ´ãåãã¦å®£ä¼ãããã¦ããã ããã°ã¨æãã¾ãã æ¸å½±ã¨è¨äºæç² çµç·¯ ç§ã¯å ã AIã«ã¤ãã¦ã¯éå¤æ¼¢ãªã®ã§ãããçæAIï¼LLMï¼ã«ã¤ãã¦ã¯ç¹ã«ChatGPTãå ¬éãããç´å¾ããèå³ããã¡ã¯ããã¾ããï¼è¾ºããã¬ãã®äººããããã¨ã«ããæãæºã ã§ããï¼ã ç¹ã«ããã°ã©ãã³ã°è¨èªå¦çã«ãããããã³ã·ã£ã«ã®é«ãã«é äºãããç¾å¨ã«è³ãã¾ã§ChatGPTï¼OpenAIï¼, Claudeï¼Anthropicï¼, Geminiï¼Googleï¼ãªã©å¤§æã®LLMã«ã¨ã©ã¾ãããã¼ã«ã«ã®LL
ç»åãAIã§çæãããã¨ãå½ããåã«ãªã£ãç¾ä»£ã ããé³æ¥½ãAIã§çæã§ãããã¨ã¯ãåãã ãããã Sunoã¨ããAIãµã¼ãã¹ã使ãã¨ãç´ äººãèãã«ã¯ããªãã¯ãªãªãã£ã®é«ã楽æ²ãçæãããã é³æ¥½ã¸ã£ã³ã«ããªãã§ãããããã¤ã³ã¹ã楽æ²ããæ¥æ¬èªã®ãã¼ã«ã«ãã楽æ²ãOKã ã å æ¥ãåå¨ããã¯ãã®ãªãAIçæã®ã·ãã£ããã楽æ²ã§è¨æ¶ãæ¹ããããã¦ãããã¨ããXã®ãã¹ããããºã£ãããçè ãèãã¦ã¿ãéãã§ã¯ãã®AIçæ楽æ²ãSunoã§çæãããã®ã ã¨æãã â²è©±é¡ã«ãªã£ãAIã·ãã£ããã çè ã¯æ®æ®µDTMã¨ããããããã½ã³ã³ã§æ²ãä½ããã¨ããã¦ããã DTMã¨ã¯ãã¹ã¯ããããã¥ã¼ã¸ãã¯ã®ç¥ã ããå製è±èªãããã ãã¨ãã¨é³æ¥½ãä½ãã®ã¯å¥½ãã§æãã®ã好ããªã®ã§ããªãªã¸ãã«ã½ã³ã°ãä½ã£ã¦æã£ã¦çºè¡¨ãã¦ããã â²ãããªæ²ãä½ã£ã¦ãã¾ã ããããæ²ãä½ãã®ã¯å¤§å¤ãªãã§ããããã®ãããã©ã æè©ããããªãã¦
æç« ãåæãã¦ã¢ã¤ãã¢åºããæèæ´çãæå©ãããGoogleã®AIãNotebookLMãã«èªåã®æ¬ãèªã¿è¾¼ã¾ããçæAIãé¸ãåºããè±ããªè¡¨ç¾ããæããã¨ããä½å®¶ãèªèº«ã®ä½é¨ã«ã¤ãã¦ã¤ã¥ãã¾ããã Fake AI âpodcastersâ are reviewing my book and itâs freaking me out - Ars Technica https://arstechnica.com/ai/2024/09/fake-ai-podcasters-are-reviewing-my-book-and-its-freaking-me-out/ ã²ã¼ã ããã¤ã³ã¹ã¤ã¼ããã«ã¾ã¤ãã3ä¸åã®æ¬ãæ¸ããã¨ããã«ã¤ã«ã»ãªã¼ã©ã³ãæ°ã¯ãèªèº«ãçæAIã®æé·ãç©æ¥µçã«è¿½ãããã¦ããã¨åç½®ããã¤ã¤ãGoogleã®æç« è¦ç´AIãNotebookLMãã®æ§è½ãããã»ã©ãã«ã¤ãã¦ããç¥ããªãã£
æ¬è¨äºã¯ä»¥ä¸ã®å ¬å¼ããã°ãåèã«ãã¦ãã¾ãã ã¯ããã« ååã®Qwen2.0ãªãªã¼ã¹ããç´3ã¶æãçµã¡ã2024å¹´9æ19æ¥ã«Qwen2.5ãæ°ãã«ãªãªã¼ã¹ããã¾ããï¼ä»åã®ãªãªã¼ã¹ã§ã¯ããã¼ã¹LLMã§ããQwen2.5ã ãã§ãªããã³ã¼ãã£ã³ã°å°ç¨ã¢ãã«ã§ããQwen2.5-Coderã¨ãæ°å¦å°ç¨ã¢ãã«ã§ããQwen2.5-Mathããªãªã¼ã¹ããã¾ããããã¹ã¦ã®ã¢ãã«ã¯ãã³ã¼ãã¼ã®ã¿ã§æ§æããã¦ããã以ä¸ã®ãããªæ§ã ãªãµã¤ãºãå©ç¨å¯è½ã§ãï¼ Qwen2.5ï¼0.5B, 1.5B, 3B, 7B, 14B, 32B, 72B Qwen2.5-Coderï¼1.5B, 7B, 32B(æºåä¸) Qwen2.5-Mathï¼1.5B, 7B, 72B ã©ã¤ã»ã³ã¹ Qwen2.5-3Bã¨Qwen2.5-72Bã¯ãããããQwen Research Licenseã¨Qwen Licenseã«ãã£ã¦ç®¡çã
NotebookLM NotebookLMã¯Googleãæä¾ãã¦ããçæAIãæ´»ç¨ãããµã¼ãã¹ã§ãããªãã¨ï¼2024å¹´10æç¾å¨ã¯ï¼ç¡æã§ä½¿ãã¾ãã NotebookLMã¯ãGoogleã®çæAI Geminiã®ãã³ã°ã³ã³ããã¹ãï¼å¤ãã®éã®ããã³ãããæ±ããã¨ãã§ããã¨ããæå³ï¼ã®ç¹å¾´ãæ´»ããããæç« ã®è¦ç´ãQA対å¿ãå¾æã§ããæç« ã¯ãç´æ¥ããã³ããã«å ¥ãã¦ããï¼ã¨æãããï¼ã®ã§ãåãªãRAGã¨ã¯éãï¼ã¨æããï¼ã¾ãã RAGã¨ãã³ã°ã³ã³ããã¹ãã®ã©ã£ã¡ãè¯ãã®ãã¨ããã®ã¯ãè²ã è°è«ãããããã§ããä¾ãã°Lost in the Middle: How Language Models Use Long Contextsã¨ããè«æã§ã¯ããã³ã°ã³ã³ããã¹ãã®çãä¸ã®æ å ±ã¯å¤±ãããããã¿ãããªãã¨ãæ¸ãã¦ããã¾ããRetrieval Augmented Generation or Long-C
Confidently evaluate, test, and ship LLM applications with a suite of observability tools to calibrate language model outputs across your dev and production lifecycle. Confidently evaluate, test, and ship LLM applications with a suite of observability tools to calibrate language model outputs across your dev and production lifecycle.
ãªãªã¼ã¹ãé害æ å ±ãªã©ã®ãµã¼ãã¹ã®ãç¥ãã
ææ°ã®äººæ°ã¨ã³ããªã¼ã®é ä¿¡
å¦çãå®è¡ä¸ã§ã
j次ã®ããã¯ãã¼ã¯
kåã®ããã¯ãã¼ã¯
lãã¨ã§èªã
eã³ã¡ã³ãä¸è¦§ãéã
oãã¼ã¸ãéã
{{#tags}}- {{label}}
{{/tags}}