¥«¥ê¥Õ¥©¥ë¥Ë¥¢½£¥¢¥ë¥Ç¥Þ¥ó¤Ë¤¢¤ëIBM¸¦µæ½ê¤¬¡¢Á°¿È¤Î¥Á¥Ã¥×¤È¹ç¤ï¤»¤ë¤È20ǯ¤«¤±¤Æ¸¦µæ¤·¤Æ¤­¤¿¤È¤¤¤¦¿·¥¿¥¤¥×¤Î¥Ç¥¸¥¿¥ë¥Á¥Ã¥×¡ÖNorthPole¡×¤Î¸ºß¤¬ÌÀ¤é¤«¤Ë¤Ê¤ê¤Þ¤·¤¿¡£³«È¯¿Ø¤Î¥À¥ë¥á¥ó¥É¥é¡¦¥â¡¼¥À»á¤Ë¤è¤ë¤È¡¢Ç¾¤Î·×»»ÊýË¡¤«¤é¥¤¥ó¥¹¥Ô¥ì¡¼¥·¥ç¥ó¤òÆÀ¤ë¤³¤È¤Ç¡¢¥³¥ó¥Ô¥å¡¼¥¿¡¼¥Á¥Ã¥×¤Î´ðËܹ½Â¤¤ò²þ¤á¤¿¤È¤Î¤³¤È¤Ç¤¹¡£

Neural inference at the frontier of energy, space, and time | Science

https://www.science.org/doi/10.1126/science.adh1174



IBM Research's new NorthPole AI chip | IBM Research Blog

https://research.ibm.com/blog/northpole-ibm-ai-chip



¡ÆMind-blowing¡Ç IBM chip speeds up AI

https://www.nature.com/articles/d41586-023-03267-0

IBM has made a new, highly efficient AI processor | Ars Technica

https://arstechnica.com/science/2023/10/ibm-has-made-a-new-highly-efficient-ai-processor/

¥³¥ó¥Ô¥å¡¼¥¿¡¼¤Î¥Á¥Ã¥×¤ÏȾƳÂΤ¬ÃÂÀ¸¤·¤Æ°ÊÍè¤Û¤ÜƱ¤¸¥Î¥¤¥Þ¥ó·¿¤Î´ðËܹ½Â¤¤ò¤·¤Æ¤¤¤Æ¡¢½èÍý¥æ¥Ë¥Ã¥È¤È½èÍý¤µ¤ì¤ë¾ðÊó¤òÊݸ¤¹¤ë¥á¥â¥ê¤ÏÊ̸ĤˤʤäƤ¤¤Þ¤¹¡£¤³¤ì¤Ë¤è¤Ã¤Æ¿ô½½Ç¯¤Ë¤ï¤¿¤ê¡¢¥·¥ó¥×¥ë¤Ê¹½Â¤¤Ç½½Ê¬¤Ê³ÈÄ¥¤¬¹Ô¤ï¤ì¤Æ¤­¤Þ¤·¤¿¤¬¡¢Âå¤ï¤ê¤Ë¡Ö¥Î¥¤¥Þ¥ó¥Ü¥È¥ë¥Í¥Ã¥¯¡×¤¬È¯À¸¤·¡¢¥á¥â¥ê¤ä¥×¥í¥»¥Ã¥µ¡¢¤½¤Î¾¤Î¥Ç¥Ð¥¤¥¹´Ö¤Ç¥Ç¡¼¥¿¤¬·Ñ³Ū¤Ë¹Ô¤­Í褹¤ë¤Ë¤¢¤¿¤Ã¤Æ»þ´Ö¤È¥¨¥Í¥ë¥®¡¼¤¬¤«¤«¤Ã¤Æ¤­¤Þ¤·¤¿¡£

¤½¤ó¤ÊÃæ¡¢IBM¤Î¥â¡¼¥À»á¤é¤Î¸¦µæ¥Á¡¼¥à¤Ï2014ǯ¤Ë¡¢Ç¾¤«¤é¥¤¥ó¥¹¥Ô¥ì¡¼¥·¥ç¥ó¤ò¼õ¤±¤¿¤È¤¤¤¦¥Á¥Ã¥×¡ÖTrueNorth¡×¤ò³«È¯.

IBM¤Î¡ÖǾ¡×¤òÌϤ·¤¿Ä¶¾ÊÅÅÎÏ¥Á¥Ã¥×¡ÖTrueNorth¡×¤¬Ãå¼Â¤Ë¿Ê²½¡¢¥Í¥º¥ß¤ÎǾ¥ì¥Ù¥ë¤ËÅþã - GIGAZINE



¤µ¤é¤Ë8ǯ¤«¤±¤Æ¡¢¥â¡¼¥À»á¤é¤Ï¥Ë¥å¡¼¥é¥ë¿äÏÀÍѤο·¥¿¥¤¥×¤Î¥Ç¥¸¥¿¥ëAI¥Á¥Ã¥×¡ÖNorthPole¡×¤Î¸¦µæ¡¦³«È¯¤ò¿Ê¤á¤Æ¤­¤Þ¤·¤¿¡£

NorthPole¤òÅëºÜ¤·¤¿¥×¥í¥È¥¿¥¤¥×¥Ç¥Ð¥¤¥¹¤Ç¡¢¿Íµ¤¤Î²èÁüǧ¼±¥â¥Ç¥ë¡ÖResNet-50¡×¤ÈʪÂθ¡½Ð¥â¥Ç¥ë¡ÖYOLOv4¡×¤Î¥Æ¥¹¥È¤ò¹Ô¤Ã¤¿¤È¤³¤í¡¢¥Õ¥ì¡¼¥à¤¢¤¿¤ê¤ËɬÍפÊÅÅÎÏ¥¸¥å¡¼¥ë¿ô¤Ç¹Í¤¨¤ë¤È¡¢¥¨¥Í¥ë¥®¡¼¸úΨ¤Ï°ìÈÌŪ¤Ê12nm¥×¥í¥»¥¹GPU¤ä14nm¥×¥í¥»¥¹GPU¤ÈÈæ³Ó¤·¤Æ25ÇܤÈÍ¥½¨¤Ç¡¢ÃÙ±ä¤â¤É¤Î¥Á¥Ã¥×¤è¤ê¤â¾¯¤Ê¤«¤Ã¤¿¤È¤Î¤³¤È¡£

NorthPole¤Î½¾ÍèÉʤȤκÇÂç¤Î°ã¤¤¤Ï¡¢256¸Ä¤Î¥³¥¢¤Ë¤½¤ì¤¾¤ìÆȼ«¤Î¥á¥â¥ê¤¬ÅëºÜ¤µ¤ì¤Æ¤¤¤ë¤È¤³¤í¤Ç¡¢´û¸¤Î¤¢¤é¤æ¤ë¥Á¥Ã¥×¤è¤ê¹â®¤ËAI¿äÏÀ¤ò¼Â¹Ô²Äǽ¤À¤½¤¦¤Ç¤¹¡£

¥â¡¼¥À»á¤Ï¡Ö¥¢¡¼¥­¥Æ¥¯¥Á¥ãŪ¤Ë¡¢NorthPole¤Ï¥³¥ó¥Ô¥å¡¼¥Æ¥£¥ó¥°¤È¥á¥â¥ê¤Î¶­³¦¤òÛ£Ëæ¤Ë¤·¤Þ¤¹¡£¸Ä¡¹¤Î¥³¥¢¤Î¥ì¥Ù¥ë¤Ç¤Ï¡¢NorthPole¤Ï¥Ë¥¢¡¦¥á¥â¥ê¡¦¥³¥ó¥Ô¥å¡¼¥Æ¥£¥ó¥°(memory-near-compute)¤Ë¸«¤¨¤Þ¤¹¤¬¡¢¥Á¥Ã¥×¤Î³°Â¦¤äÆþ½ÐÎϤΥì¥Ù¥ë¤«¤é¤Ï¥¢¥¯¥Æ¥£¥Ö¥á¥â¥ê¤Ë¸«¤¨¤Þ¤¹¡×¤ÈÀâÌÀ¤·¤Æ¤¤¤Þ¤¹¡£

¤¿¤À¤·¡¢NorthPole¤ÎºÇÂç¤ÎÍøÅÀ¤ÏƱ»þ¤ËÀ©Ìó¤Ç¤â¤¢¤ê¡¢¥ª¥ó¥Ü¡¼¥É¥á¥â¥ê¤«¤é¤·¤«´Êñ¤Ë°ú¤­½Ð¤»¤Ê¤¤¤È¤¤¤¦ÌäÂ꤬¤¢¤ë¤È¤Î¤³¤È¡£³Æ¥³¥¢¤Ë¤Ï½½Ê¬¤Ê¥á¥â¥ê¤òÅëºÜ¤·¤Æ¤¤¤ë¤â¤Î¤Î¡¢¤³¤Î¥á¥â¥ê¤Ï¡Ö²¿¤Ç¤â²°¡×¤È¤·¤Æ¤ÎÆ°ºî¤òÁÛÄꤷ¤¿¤â¤Î¤Ç¤Ï¤Ê¤¯¡¢¡Ö¤¢¤¯¤Þ¤ÇNorthPole¤Ï¿äÏÀÀìÍѤǤ¹¡×¤È¥â¡¼¥À»á¤Ï½Ò¤Ù¤Æ¤¤¤Þ¤¹¡£