°æÁ¦¿Í »ê¾÷À» µÚÈçµå´Â ÀΰøÁö´ÉÀÇ ÁøÈ­

åǥÁö

°æÁ¦ ÀÎÇÁ¶óÀÇ Áß¿äÇÑ ¿ä¼Ò·Î ¶°¿À¸¥ ÀΰøÁö´ÉAIÀÇ º¯Ãµ»ç´Â ´ÜÀϼ¼Æ÷ À¯±âüµéÀÌ °ø»ý Ä¿¹Â´ÏƼ¸¦ ÀÌ·ï Áö¹èÇÏ´ø 38¾ï ³â ÀüÀÇ »ýÅ°è¿Í À¯»çÇÑ Ãø¸éÀÌ ÀÖ´Ù. ±×¸®°í ÀÌÁ¦ ¿ì¸®´Â įºê¸®¾Æ..





°æÁ¦ ÀÎÇÁ¶óÀÇ Áß¿äÇÑ ¿ä¼Ò·Î ¶°¿À¸¥ ÀΰøÁö´ÉAIÀÇ º¯Ãµ»ç´Â ´ÜÀϼ¼Æ÷ À¯±âüµéÀÌ °ø»ý Ä¿¹Â´ÏƼ¸¦ ÀÌ·ï Áö¹èÇÏ´ø 38¾ï ³â ÀüÀÇ »ýÅ°è¿Í À¯»çÇÑ Ãø¸éÀÌ ÀÖ´Ù. ±×¸®°í ÀÌÁ¦ ¿ì¸®´Â įºê¸®¾Æ±âÀÇ ´ëÆø¹ßó·³ °Å´ëÇÑ º¯È­¸¦ ¸ô°í ¿Ã ÀΰøÁö´ÉÀÇ Æø¹ßÀû ÁøÈ­¸¦ ´«¾Õ¿¡ µÎ°í ÀÖ´Ù. ÀΰøÁö´ÉÀÌ ºñÁî´Ï½º »ýÅ°迡 ¹ÌÄ¥ ¿µÇâ°ú ±×·Î ÀÎÇØ ¹ß»ýÇÒ À§±â¿Í ±âȸ´Â ¹«¾ùÀΰ¡?


ÀΰøÁö´É°ú ¾÷¹« ÀÚµ¿È­°¡ Áö´Ñ ÀáÀçÀû ¿µÇâÀº ¿ì¸®°¡ ÀÍÈ÷ ¾Ë°í ÀÖ´Â ÁÖÁ¦´Ù. ¹«ÀÎ ÀÚµ¿Â÷, ÀÚµ¿ÁÖÇà ÀÚµ¿Â÷µµ ÀΰøÁö´ÉÀÇ ¹ßÀü ´ö¿¡ °¡´ÉÇØÁø °ÍÀÌ´Ù. ÀÌÁ¦ ÀΰøÁö´ÉÀÌ ±â¾÷ÀÇ ºñÁî´Ï½º¿¡ ¹ÌÄ¥ ¿µÇâ¿¡ ´ëÇØ »ìÆ캸µµ·Ï ÇÏÀÚ.


ÀΰøÁö´ÉÀº Åë»ó ¡°ÄÄÇ»ÅÍ ¼ÒÇÁÆ®¿þ¾îÀÇ ÀÇ»ç °áÁ¤, ³íÁõ, ¹®Á¦ ÇØ°áÀ» ¿¬±¸ÇÏ´Â ÄÄÇ»ÅÍ °úÇÐÀÇ ÇÑ ºÐ¾ß¡±·Î Á¤ÀǵȴÙ. ÇÏÁö¸¸ ÀÌ´Â Àü¹®¿ë¾î·Î ¡®°­ÇÑstrong AI¡¯ÀÇ Ãø¸éÀ» Á¤ÀÇÇÑ °ÍÀ¸·Î, Àΰ£ÀÇ µÎ³úó·³ À¶Å뼺À» ¹ßÈÖÇÒ ¼ö ÀÖ´Â ÀΰøÁö´É ½Ã½ºÅÛ¿¡ ´ëÇÑ ¼³¸íÀ̶ó°í ÇÒ ¼ö ÀÖ´Ù. ±×·± ÀΰøÁö´ÉÀÌ ½ÇÇöµÇ·Á¸é ¾ÕÀ¸·Î ÃÖ¼ÒÇÑ 30³âÀº ´õ ±â´Ù·Á¾ß ÇÒ °ÍÀÌ´Ù.


´ë½Å ¿À´Ã³¯ ÇÏ·ç°¡ ¸Ö´Ù ÇÏ°í ÀÀ¿ëµÇ¾î µîÀåÇÏ´Â °ÍÀº ¡®¾àÇÑweak AI¡¯´Ù. ¾àÇÑ ÀΰøÁö´ÉÀº ¸Å¿ì ÁýÁßÀ» ¿äÇÏ´Â ±â¼ú ºÐ¾ß¿¡¼­ »ç¿ëµÇ´Âµ¥, Çϳª ÀÌ»óÀÇ ´Ü¼øÇÑ ¾÷¹«¸¦ Àΰ£º¸´Ù ´õ È¿À²ÀûÀ¸·Î ¼öÇàÇÑ´Ù. ´ÙÀ½°ú °°Àº ¿¹¸¦ µé ¼ö ÀÖ´Ù.


¢º Ç×°ø±âÀÇ ºñÇà °èȹÀ» °áÁ¤ÇÏ°í ÃÖÀûÀÇ Âø·ú °ÔÀÌÆ®¸¦ ¼±ÅÃÇÏ´Â Ç×°ø±³Åë°üÁ¦ ½Ã½ºÅÛ


¢º UPS °°Àº ¹è¼Û¾÷üµéÀÌ ½Ã°£°ú ¿¬·á¸¦ Àý¾àÇϵµ·Ï Æ®·° °æ·Î¸¦ ÁöÁ¤ÇØ ÁÖ´Â ¹°·ù ¾ÖÇø®ÄÉÀ̼Ç


¢º ´ãº¸ ´ëÃâ ½ÅûÀÚµéÀÇ ½Å¿ëµµ¸¦ Æò°¡ÇÏ´Â ´ëÃâ ó¸® ½Ã½ºÅÛ


¢º °É·Á¿À´Â ÀüÈ­¸¦ ó¸®ÇÏ°í ÀÚµ¿È­µÈ °í°´ ¼­ºñ½º¸¦ Á¦°øÇÏ´Â À½¼ºÀÎ½Ä ÀüÈ­±â


¢º ´Ù¾çÇÑ µ¥ÀÌÅÍ ¼Ò½º¸¦ °Ë»öÇÏ°í °£´ÜÇÑ ´äÀ» ÇÒ ¼ö ÀÖ´Â, ¾ÖÇÃÀÇ ½Ã¸®Siri¿Í °°Àº µðÁöÅÐ °³Àκñ¼­


ÀÌ ¸ðµç °æ¿ì¿¡¼­ ÀΰøÁö´ÉÀº Àΰø½Å°æ¸ÁNeural NetworksÀÌ Á¤º¸¸¦ ó¸®ÇÏ°í ´äÀ» »êÃâÇÏ´Â µ¥ »ç¿ëÇÏ´Â ÀÏ·ÃÀÇ ¾Ë°í¸®Áò(°ø½ÄÀ̳ª ±ÔÄ¢µé)À» Åä´ë·Î ÇÏ°í ÀÖ´Ù. ¼ö½Ê ³â µ¿¾È °úÇÐÀÚµéÀÌ ÀΰøÁö´É °³¹ßÀ» ¿¬±¸ÇØ ¿ÔÁö¸¸, ÀÌÁ¦¾ß Áß¿äÇÑ ºñÁî´Ï½º µµ±¸·Î µîÀåÇÑ ÀÌÀ¯´Â ´ÙÀ½ µÎ °¡ÁöÀÇ Áß¿äÇÑ ¹ßÀü ¶§¹®ÀÌ´Ù.


1. ó¸® ´É·ÂÀÌ °¡¼ÓÈ­µÇ°í ÀÖ´Ù. ÀÎÅÚ Ã¢¸³ÀÚ °íµç ¹«¾îGordon MooreÀÇ À̸§À» µý ¡®¹«¾îÀÇ ¹ýÄ¢Moore¡¯s Law¡¯ÀÌ ¿¹ÃøÇÑ ¹Ù¿Í °°ÀÌ, Áö³­ 40³â µ¿¾È Ĩ ÇÑ °³´ç Æ®·£Áö½ºÅÍÀÇ ¼ö´Â ¸Å³â 2¹è°¡·® Áõ°¡ÇØ ¿Ô´Ù. ¿£ºñµð¾ÆNvidia µîÀÇ ±â¾÷Àº ¼º´ÉÀ» Çâ»ó½ÃÅ°±â À§ÇØ Ä¨ ¾È¿¡ Áß¾Óó¸®ÀåÄ¡CPU ÄÚ¾î ¿Ü¿¡ ±×·¡ÇÈó¸®ÀåÄ¡GPU Äھ Ãß°¡ÇÑ´Ù. CPU´Â ÇÑ ¹ø¿¡ ÇÑ °¡Áö ¾÷¹«¸¦ ¿¬¼ÓÀûÀ¸·Î ¼öÇàÇÏ´Â Á÷·Ä 󸮹æ½ÄÀÇ ¸î¸î ÄÚ¾î·Î ±¸¼ºµÇ´Â ¹Ý¸é, GPU´Â µ¿½Ã¿¡ ¿©·¯ ¾÷¹«¸¦ ¼öÇàÇÏ´Â º´·Ä 󸮹æ½ÄÀÇ ¼öõ °³ ÄÚ¾î·Î ÀÌ·ïÁ® ÀÖ´Ù. ÀÌ´Â °ð Àΰø½Å°æ¸ÁÀÌ ¹æ´ëÇÑ ¾çÀÇ µ¥ÀÌÅ͸¦ µ¿½Ã´Ù¹ßÀûÀ¸·Î ºÐ¼®ÇÏ´Â ¼ÒÀ§ ¡®µö ·¯´×deep learning¡¯ÀÌ °¡´ÉÇÏ´Ù´Â Àǹ̴Ù.


2. µ¥ÀÌÅÍÀÇ ¾ç°ú ÀúÀå °¡´É ¿ë·®ÀÌ Áö¼ÓÀûÀ¸·Î È®´ëµÇ°í ÀÖ´Ù. ¿¬±¸±â°ü ½ÅÅ×ÇÁSINTEF¿¡ µû¸£¸é, ¡°ÇöÀç Àü ¼¼°è ¸ðµç µ¥ÀÌÅÍ Áß 90%°¡ Áö³­ 2³â µ¿¾È »ý¼º¡±µÇ¾ú´Ù°í ÇÑ´Ù. ±×¸®°í ½ÃÀåÁ¶»ç ±â°ü IDCÀÇ º¸°í¼­¿¡ µû¸£¸é ¸Å³â »ý¼ºµÇ°í º¹Á¦µÇ´Â µ¥ÀÌÅÍÀÇ ¾çÀº 2³â¸¶´Ù µÎ ¹è°¡ µÇ¾î, 2020³â¿¡´Â 44Á¦Å¸¹ÙÀÌÆ®(44Á¶ ±â°¡¹ÙÀÌÆ®)¿¡ À̸¦ °ÍÀ̶ó°í ÇÑ´Ù. ÇÑÆí Çϵåµð½ºÅ© Àü¹®±â¾÷ ½Ã°ÔÀÌÆ®SeagateÀÇ ÀüÀÓ ÃÖ°í±â¼úÃ¥ÀÓÀÚ ¸¶Å© Å©¶óÀÌ´õMark KryderÀÇ À̸§À» µý ¡®Å©¶óÀÌ´õÀÇ ¹ýÄ¢Kryder¡¯s Law¡¯¿¡ ÀÇÇϸé Çϵåµð½ºÅ© µå¶óÀ̺êÀÇ ÀúÀå ¿ë·®Àº 18°³¿ù¸¶´Ù µÎ ¹è°¡ µÈ´Ù. Áï µ¥ÀÌÅÍ ÀúÀå ºñ¿ëÀÌ 18°³¿ù¸¶´Ù Àý¹ÝÀ¸·Î ÁÙ¾îµç´Ù´Â ¸»°ú °°´Ù. ÀÌ´Â Àΰø½Å°æ¸ÁÀ» ¡®°¡¸£Ä¡´Â¡¯ µ¥ ÇÊ¿äÇÑ ¸ðµç µ¥ÀÌÅ͸¦ Àú·ÅÇÏ°Ô ÀúÀåÇÏ°í, Àΰø½Å°æ¸ÁÀÌ ºÐ¼®ÇÒ ¸ÅÀÏ°°ÀÌ »ý¼ºµÇ´Â ¾î¸¶¾î¸¶ÇÑ ¾çÀÇ »õ·Î¿î µ¥ÀÌÅ͸¦ ÃëÇÕÇÏ´Â °ÍÀÌ °¡´ÉÇØÁø´Ù´Â Àǹ̴Ù.


µ¥ÀÌÅÍ¿Í Ã³¸®´É·ÂÀÌ ÀÌ·¸µí ¾öû³ª°Ô Æø¹ßÀûÀ¸·Î Áõ°¡ÇÏ°í ¹ßÀüÇÑ °á°ú´Â, ½Å¿ëÄ«µå ¹üÁ˸¦ Àâ¾Æ³»°Å³ª ƯÁ¤ °í°´¿¡°Ô ³ëÃâ½Ãų ±¤°í¸¦ °áÁ¤Çϱâ À§ÇØ ¿Â¶óÀÎ ±¸¸Å ÆÐÅÏÀ» ºÐ¼®ÇÏ´Â ºñÁî´Ï½º ÀÀ¿ë ºÐ¾ß¿¡¼­ ÀÌ¹Ì Ã£¾Æº¼ ¼ö ÀÖ´Ù. »õ·Î¿î ±â´ÉµéÀÌ º¸´Ù ¿Ïº®ÇÑ ¸ð½ÀÀ» °®Ãç°¨¿¡ µû¶ó ÀÌ ±â¼úÀÇ Àû¿ë °¡´É¼ºµµ È®´ëµÉ Àü¸ÁÀÌ´Ù. ÀÏ·Ê·Î ±¸±Û°ú ½ºÅÄÆÛµå ´ëÇб³ÀÇ °úÇÐÀÚµéÀº ÀΰøÁö´É¸Á, ÄÄÇ»ÅÍ ºñÀüComputer Vision, ÀÚ¿¬¾ð¾î 󸮰úÁ¤À» °áÇÕÇÏ´Â ½Ã½ºÅÛÀ» °³¹ß Áß¿¡ ÀÖ´Ù.


À§ µÎ °¡Áö ¹ßÀüÀÌ °¡Á®¿Â °á°ú°¡ À̹ÌÁö¸¦ ÅؽºÆ®·Î º¯È¯ÇÒ ¼ö ÀÖ´Â ÀΰøÁö´É ½Ã½ºÅÛÀÌ´Ù. ¿¹¸¦ µé¸é, ÀÌ ½Ã½ºÅÛÀº »çÁø¿¡¼­ ¹«½¼ ÀÏÀÌ ÀϾ´ÂÁö Á¤È®ÇÏ°Ô ÆľÇÇØ ¡°ºñÆ÷Àåµµ·Î¿¡¼­ ¿ÀÅä¹ÙÀ̸¦ Ÿ°í ÀÖ´Â »ç¶÷¡±À̶ó´Â ¼³¸íÀ» ³»³õ´Â °ÍÀÌ °¡´ÉÇÏ´Ù.


»çÁø°ú ¿µ»óÀº ÀÌÁ¦±îÁö ÀΰøÁö´É ½Ã½ºÅÛÀÌ Çؼ®ÇÏ´Â µ¥ ¾î·Á¿òÀ» °Þ¾î¿Ô´ø ºñÁ¤Çü µ¥ÀÌÅÍÀÇ ÀÏÁ¾ÀÌ´Ù. ¿ÃÇظ¸ Çصµ 1Á¶¿¡ ´ÞÇÏ´Â »çÁøµéÀÌ (¾Æ¸¶ ½º¸¶Æ®Æù ´öÀÌ°ÚÁö¸¸) ÂïÈ÷°í ±× Áß ¼ö½Ê¾ï ÀåÀÌ ¿Â¶óÀο¡ °Ô½ÃµÇ¸®¶ó°í °¡Á¤ÇØ º¸ÀÚ. ÀÌ »çÁøµéÀ» È°¿ëÇÒ ¼ö ÀÖ´Â ÀÀ¿ë ºÐ¾ß Áß Çϳª´Â ÀΰøÁö´ÉÀ» ÀÌ¿ëÇÑ º¸¾È ¾÷¹«°¡ µÉ ¼ö ÀÖ´Ù. º¸½ºÅÏ ¸¶¶óÅæ Æø¹ß »ç°ÇÀÇ ¿ëÀÇÀÚ¸¦ ½Äº°Çϱâ À§ÇØ FBI°¡ ¿ëÀÇÀÚÀÇ »çÁø°ú ¿µ»óÀ» »ç¿ëÇß´ø ÀÏÀ» ¶°¿Ã·Á º¸¶ó. ÀÌÁ¦ µ¥ÀÌÅÍÀÇ ¾çÀÌ ¼öÁ¶¿¡ ´ÞÇÏ´Ï ±×¸¸Å­ ¹üÁËÀÚ¿Í Å×·¯¸®½ºÆ®µéÀÇ È°µ¿À» °¨½ÃÇÒ ±âȸ°¡ È®´ëµÈ´Ù°í ÇÒ ¼ö ÀÖ´Ù.


·Îº¿°øÇÐÀº ÀΰøÁö´ÉÀÇ ¹ß´Þ·Î ÇýÅÃÀ» ÀÔ°Ô µÉ ¶Ç ´Ù¸¥ ÀÀ¿ë ºÐ¾ß´Ù. ¿¹¸¦ µéÀÚ¸é, ±¸±Û°ú ½ºÅÄÆÛµåÀÇ ¿¬±¸°á°ú´Â ÈÞ¸Ó³ëÀÌµå ·Îº¿°ú ÀÚµ¿ÁÖÇà ÀÚµ¿Â÷°¡ ÁÖÀ§¿¡ ÀÖ´Â ¹°Ã¼¸¦ ÆľÇÇÏ°í, ÀÌ¿¡ ´ëóÇÏ´Â ¹æ¹ýÀ» °áÁ¤ÇÏ°í, À̸¦ »ç¶÷µé¿¡°Ô ¼³¸íÇÒ ¼ö ÀÖ°Ô ÇØÁÙ °ÍÀÌ´Ù. ¼Ò¸Å¾÷üµéÀº ºñµð¿ÀÄ«¸Þ¶ó¿Í °áÇÕÇÑ ±â¼úÀ» È°¿ëÇؼ­ ¸ÅÀå ³»ÀÇ ¼îÇΰ´À» ÃßÀûÇÒ ¼ö ÀÖ´Ù. µµ³­À» ¹æÁöÇÒ »Ó¸¸ ¾Æ´Ï¶ó ¸ÂÃã »óÇ°À» ¼±º¸ÀÌ°í °í°´ÀÇ ½º¸¶Æ®ÆùÀ¸·Î ÇÒÀÎ ÄíÆùÀ» Àü¼ÛÇÏ´Â µî ÆÇÃË È°µ¿¿¡ È°¿ëÇÒ ¼ö ÀÖÀ» °ÍÀÌ´Ù.


ÀÌ·¯ÇÑ °úÇÐÀû ÁøÈ­¿¡ µû¶ó ¿ì¸®´Â ÀÌ ºÐ¾ß¿¡¼­ ÇâÈÄ ´ÙÀ½°ú °°Àº ¿¹ÃøÀ» ³»·Á º»´Ù.


ù°, ÀΰøÁö´ÉÀÌ ¹ßÀüÇÔ¿¡ µû¶ó ÀΰøÁö´É ½Ã½ºÅÛÀº ºñÁî´Ï½º¿¡¼­ ´õ¿í Æø³Ð°Ô Àû¿ëµÉ °ÍÀÌ´Ù.


¾Ë°í¸®ÁòÀÇ °³¼± ±×¸®°í °è»ê ´É·Â°ú µ¥ÀÌÅÍ ÀúÀå ¿ë·®ÀÇ Áö¼ÓÀû ¹ßÀüÀº, ÀΰøÁö´ÉÀÇ À§»óÀ» ¡®¼±µµ±â¾÷ÀÌ »ç¿ëÇÏ´Â µµ±¸¡¯¿¡¼­ ¿À´Ã³¯ »ì¾Æ³²±â À§ÇØ ¿ä±¸µÇ´Â ¡®°æÀï ¿ìÀ§ÀÇ È®º¸ ¼ö´Ü¡¯À¸·Î ³ô¿©ÁÙ °ÍÀÌ´Ù. °¡Æ®³ÊÀÇ ¿¬±¸¿¡ µû¸£¸é, ÀΰøÁö´ÉÀº ±â¾÷ °æ¿µ¿¡ ÇÊ¿äÇÑ ºñ¿ëÀ» 2019³âÀ̸é 30% ÁÙ¿©ÁÙ °ÍÀ̶ó°í ÇÑ´Ù. ÀÌ·± ½Ã·ù¿¡¼­ µÚóÁö´Â ±â¾÷µéÀº °í°´µéÀÌ ±â´ëÇÏ´Â °³ÀÎ ¸ÂÃãÇü Àúºñ¿ë Á¦Ç°°ú ¼­ºñ½º¸¦ Àü´ÞÇÒ ¼ö ¾øÀ» °ÍÀÌ´Ù. ÀÌ´Â ÀÎÅͳݰú »ó´çÈ÷ À¯»çÇÑ ÆÐÅÏÀÌ´Ù. ºÎ°¡ÀûÀÎ ¸¶ÄÉÆà ä³Î¿¡ ºÒ°úÇß´ø ÀÎÅͳÝÀº Áö³­ 30³â µ¿¾È ÁøÈ­ÇÏ¿© ¾÷°èÀÇ °¡Ä¡ üÀÎÀ» ÀçÇü¼ºÇÏ°í ¿ÏÀüÈ÷ »õ·Î¿î ºñÁî´Ï½º ¸ðµ¨À» °¡´ÉÇÏ°Ô ÇÏ´Â °­·ÂÇÏ°í Æı«ÀûÀÎ ¿øµ¿·ÂÀ¸·Î ¹ßÀüÇß´Ù.


µÑ°, ÀΰøÁö´ÉÀº Áö½Ä ±Ù·ÎÀÚµéÀÇ »ý»ê¼º°ú Á¤È®¼ºÀ» Çâ»ó½Ãų °ÍÀÌ´Ù.


<Å×Å©¸®ÆÛºí¸¯TechRepublic>¿¡ µû¸£¸é, ĵ¼îKensho¶ó´Â ½Å»ý ±â¾÷Àº ¡®±ÝÀ¶¾÷°è¸¦ À§ÇÑ ¼¼°è ÃÖÃÊÀÇ Àü»ê Áö½Ä ¿£Áø¡¯À» Ç¥¹æÇϴµ¥, ¹æ´ëÇÑ º´·Ä Åë°è ¿¬»ê, ÀÚ¿¬¾ð¾î ÀÔ·Â, ºòµ¥ÀÌÅÍ, ¸Ó½Å ·¯´×À» »ç¿ëÇؼ­ º¹ÀâÇÑ ±ÝÀ¶ °ü·Ã Áú¹®À» ½±°Ô ¼³¸íÇÏ´Â ½Ã½ºÅÛÀÌ´Ù. ÀÌ´Â ¿ù½ºÆ®¸®Æ®, ·±´ø µî ¿©·¯ ±ÝÀ¶ Áß½ÉÁö¿¡¼­ ½ÃÀåÀÇ ¿ªµ¿¼ºÀ» ºÐ¼®Çϱâ À§ÇØ °í¿ëµÈ °ªºñ½Ñ ¡®±ÝÀ¶½ÃÀå ºÐ¼®°¡¡¯µéÀÇ °í¿ë ¾ÈÀü¿¡ ÀáÀçÀûÀÎ À§ÇùÀÌ µÉ ¸¸ÇÑ °³¹ßÀÌ´Ù. ÇÏÁö¸¸ Àû¾îµµ °¡±î¿î ¹Ì·¡¿¡ ÀΰøÁö´É ½Ã½ºÅÛÀÌ °íµµ·Î ¼÷·ÃµÈ Àΰ£ ³ëµ¿ÀÚ¸¦ ´ë½ÅÇÏÁö´Â ¾ÊÀ» °¡´É¼ºÀÌ ´ÙºÐÇÏ´Ù. ±×º¸´Ù´Â Àü¹®°¡µé°ú ÇÔ²² ¿î¿µµÇ¸é¼­ ±×µéÀÇ ´É·üÀ» Çâ»ó½ÃÅ°´Â ¿ªÇÒÀ» ÇÒ °ÍÀÌ´Ù.


¼Â°, ÀÇ·á »ê¾÷¿¡¼­ ÀΰøÁö´ÉÀº Áúº´ÀÇ ¿¹¹æ ¹× Ä¡·á¿¡ ÀÏ´ë Çõ¸íÀ» °¡Á®¿Ã °ÍÀÌ´Ù.


IBMÀÇ Á¾¾ç Áø´Ü ½Ã½ºÅÛ ¿Ó½¼Watson for OncologyÀ» »ý°¢ÇØ º¸¶ó. ÀÌ ½Ã½ºÅÛÀº ȯÀÚÀÇ Áõ»óÀ» ºÐ¼®Çϱâ À§ÇØ ¿©·¯ ÀÇ·á µ¥ÀÌÅͺ£À̽º¿¡ ÀúÀåµÈ ¸ðµç Áö½ÄÀ» È°¿ëÇÏ¿©, Ä¡·á¹ý ¸®½ºÆ®(È¿°ú È®·üÀÌ ³ôÀº ¼ø¼­´ë·Î)¿Í ÇÔ²² Âü°íÇÒ ¹®Çå ¸ñ·ÏÀ» Á¦°øÇÑ´Ù. ¶Ç °¡Æ®³ÊÀÇ ¿¬±¸¿¡ µû¸£¸é, ¹«¼±À¸·Î °Ç°­ »óŸ¦ ¸ð´ÏÅ͸µÇÏ´Â ±â¼ú ´öºÐ¿¡ 2020³â¿¡´Â ¼±Áø±¹¿¡¼­ÀÇ ±â´ë ¼ö¸íÀÌ Æò±Õ 6°³¿ù Á¤µµ ´õ ´Ã¾î³¯ °ÍÀ̶ó°í ÇÑ´Ù.


³Ý°, ÀΰøÁö´ÉÀº ¾Ë°í¸®ÁòÀ» ÅëÇØ ÀÚµ¿È­µÉ ¼ö ÀÖ´Â ´Ù¾çÇÑ ¾÷¹«¸¦ ´ã´çÇÏ°ÚÁö¸¸, Àΰ£ °íÀ¯ÀÇ ÀÚÁúÀ» ¸ð¹æÇÏÁö´Â ¸øÇÒ °ÍÀÌ´Ù.


¿¹¸¦ µé¾î, ¿Ó½¼ ½Ã½ºÅÛÀÌ È¯ÀÚÀÇ ¾ÏÀ» Áø´ÜÇÏ°í ÃÖ¼±ÀÇ Ä¡·á ¹æ¹ýÀ» µµÃâÇÏ´Â ÀÏ¿¡ À־´Â Àΰ£ Àǻ纸´Ù ³ªÀ»Áö ¸ô¶óµµ, ȯÀÚµéÀº ¾Æ¸¶ ±â°èº¸´Ù´Â °°Àº Àΰ£À¸·ÎºÎÅÍ Ä¡·á¿¡ ´ëÇÑ ¼³¸íÀ» µè±æ ¹Ù¶ö °ÍÀÌ´Ù. Áø·á½Ç¿¡¼­ºÎÅÍ ¹Ì¿ë½Ç°ú ÆǸŠ¸ÅÀå¿¡ À̸£±â±îÁö ÀΰøÁö´ÉÀº ÀÌ·± ȯ°æµéÀÇ ÇÁ·Î¼¼½º¸¦ º¸´Ù È¿À²È­½ÃÅ°°ÚÁö¸¸, °í°´µé°ú °³ÀÎÀûÀÌ°í °¨Á¤ÀûÀ¸·Î ¼ÒÅëÇϱâ À§Çؼ­´Â Àΰ£ ±Ù·ÎÀÚµéÀÌ ¿©ÀüÈ÷ ÇÊ¿äÇÒ °ÍÀÌ´Ù.


´Ù¼¸Â°, ÀΰøÁö´ÉÀº °æÁ¦¸¦ ¹Ù²ã³õÀ» ±â¼úµéÀ» °¡´ÉÇÏ°Ô ÇÒ °ÍÀ̸ç, ¶ÇÇÑ ±×·ÎºÎÅÍ ÀΰøÁö´Éµµ µµ¿òÀ» ¾òÀ» °ÍÀÌ´Ù.


20¼¼±â ÃÊ ÀÚµ¿Â÷, µµ·Î, ¼®À¯ »ê¾÷, ±×¸®°í ´ë·®»ý»ê µîÀÇ ±â¼úÀÌ ÇÔ²² ¹ßÀüÇϸ鼭 ¹®¸íÀ» º¯È­½ÃŲ °Í°ú ¸¶Âù°¡Áö·Î, ÀΰøÁö´ÉÀº µ¶ÀÚÀûÀ¸·Î´Â »ýÁ¸ÀÌ ºÒ°¡´ÉÇÑ ¹«¼öÈ÷ ¸¹Àº »óÈ£ ÀÇÁ¸Àû »ê¾÷°ú ±â¼úÀÇ °áÇÕü Áß ÇÑ ºÎºÐÀÌ´Ù. ÀΰøÁö´ÉÀÇ ÀáÀç·ÂÀ» ÃÖ´ëÇÑ À̲ø¾î³»´Â ÀÏÀº ³×Æ®¿öÅ©, ¹ÙÀÌ¿ÀĨ, ¾çÀÚ ÄÄÇ»ÆÃ, ¿þ¾î·¯ºí ÄÄÇ»ÅÍ, ·Îº¿°øÇÐ, À½¼ºÀνÄ, ¸Ó½Å ·¯´×, »ç¹°ÀÎÅͳÝÀÇ È°¼ºÈ­¿¡ ´Þ·Á ÀÖÀ¸¸ç, ¶ÇÇÑ ÀÌ ¸ðµç ±â¼úµéÀº ÀΰøÁö´É¿¡ ÀÇÁ¸ÇÏ°í Àֱ⵵ ÇÏ´Ù.


¿©¼¸Â°, ÀΰøÁö´ÉÀÌ Á¦´ë·Î °ü¸®µÇÁö ¾Ê´Â´Ù¸é, ÀÌ·ÐÀûÀ¸·Î´Â ¹®¸íÀÇ Á¾¸»·Î À̾îÁú Å« À§ÇèÀ» ¾È°í ÀÖ´Ù.


¡®°­ÇÑ AI¡¯´Â ¸Ó½Å ·¯´×À» ÅëÇØ Áøº¸ÇØ °¥ °ÍÀÌ°í ȯ°æ¿¡ ´ëÀÀÇØ ½º½º·Î¸¦ ¼öÁ¤ÇÏ¸ç °è¼Ó ¶È¶ÈÇØÁú °ÍÀ̱⠶§¹®¿¡, Àΰ£ÀÇ ÅëÁ¦¸¦ ¹þ¾î³ª ¹ßÀüÇÒ °¡´É¼ºÀÌ ÀÖ´Ù. ÇöÀç ÀΰøÁö´É ½Ã½ºÅÛÀÇ ¾Ë°í¸®Áò¿¡´Â ºÎÁ¤Àû °á°ú¸¦ ¹æÁöÇϱâ À§ÇÑ ±ÔÄ¢µéÀÌ Æ÷ÇԵŠÀÖ´Ù. ¿¹¸¦ µé¾î, ·Îº¿ Áø°øû¼Ò±âÀÇ ¾÷¹«°¡ ¡°Áý¾È ¸ÕÁö¸¦ °¡´ÉÇÑ ÇÑ ¸¹ÀÌ ¸ðÀ¸±â¡±·Î ±ÔÁ¤µÇ¾î ÀÖ´Ù¸é, °è¼ÓÇؼ­ ¸ÕÁö¸¦ ¸ðÀ¸±â À§ÇØ ¿©Å²¯ ¸ðÀº ¸ÕÁö¸¦ ´Ù½Ã ¹Ù´Ú¿¡ ¹ö·È´Ù ¸ðÀ¸±â¸¦ ¹Ýº¹ÇÒ °ÍÀÌ´Ù. ±×·¯´Ï ±× ´ë½Å ¡°¹Ù´ÚÀ» ±ú²ýÇÏ°Ô À¯ÁöÇϱ⡱ó·³ ÇÒ ÀÏÀ» ½ÅÁßÈ÷ ÀÔ·ÂÇØ¾ß ÇÑ´Ù. ¸¶Âù°¡Áö·Î ÀΰøÁö´É ½Ã½ºÅÛ¿¡ ÀÎÅͳÝÀÇ ¸ðµç Áö½ÄÀ» Èí¼öÇؼ­ ¡°Áö±¸ ÀÚ¿øÀ» º¸Á¸ÇÒ ¹æ¹ýÀ» ã¾Æ¶ó.¡±´Â ¸í·ÉÀ» ÀÔ·ÂÇß´Ù¸é, ÃÖ¼±ÀÇ ÇØ°áÃ¥Àº Àηù¸¦ ¸ê¸Á½ÃÅ°´Â °ÍÀ̶õ °áÁ¤À» ³»¸± ¼öµµ ÀÖ´Ù. Å͹«´Ï¾ø´Â ¾ê±â·Î µé¸± ¼ö ÀÖ°ÚÁö¸¸ ½ºÆ¼ºì ȣŷ, ºô °ÔÀÌÃ÷, ¿¤·Ð ¸Ó½ºÅ©µµ ÀΰøÁö´ÉÀÇ À§Ç輺¿¡ ´ëÇØ ¿ì·Á¸¦ Ç¥ÇØ ¿Ô´Ù. ½ºÄ«ÀÌÇÁSkype¸¦ â¾÷ÇÑ ¾á ŸÀϸ°Jaan TallinnÀº ÃÖ±Ù ¡®»îÀÇ ¹Ì·¡ ¿¬±¸¼Ò¡¯¶ó´Â Á¶Á÷À» â¼³Çß´Ù. ¼¼°èÀûÀ¸·Î ¶Ù¾î³­ ÀΰøÁö´É Àü¹®°¡µéÀÌ ´Ù¼ö Æ÷ÇÔµÈ ÀÚ¹®À§¿øȸ¿Í ¿¤·Ð ¸Ó½ºÅ©ÀÇ 1,000¸¸ ´Þ·¯ ±âºÎ±ÝÀ¸·Î ¹ßÁ·ÇÑ ÀÌ ¿¬±¸¼ÒÀÇ ÀÓ¹«´Â ±â¼úÀÌ ´ëÀç¾ÓÀ» ÀÏÀ¸Å°Áö ¸øÇϵµ·Ï ÇÒ ¹æ¹ýÀ» ¸ð»öÇÏ´Â ÀÏÀ̶ó°í ÇÑ´Ù.


* *

References List :
1. For more information about GPU-accelerated computing, visit the Nvidia website at:
http://www.nvidia.com/object/what-is-gpu-computing.html


2. For more information about ¡°deep learning¡± using neural networks, visit the Nvidia website at:
https://developer.nvidia.com/deep-learning


3. For more information about data expansion, visit the SINTEF website at:
http://www.sintef.no/en/corporate-news/big-data?for-better-or-worse/


4. ComputerWeekly, April 9, 2014, ¡°Data Set to Grow 10-Fold by 2020 as Internet of Things Takes Off,¡± by Antony Adshead. ¨Ï 2014 TechTarget, Inc. All rights reserved.
http://www.computerweekly.com/news/2240217788/Data-set-to-grow-10-fold-by-2020-as-internet-of-things-takes-off


5. Scientific American, August 2005, ¡°Kryder¡¯s Law,¡± by Chip Walter. ¨Ï 2005 Scientific American, a division of Nature America, Inc. All rights reserved.
http://www.scientificamerican.com/article/kryders-law/


6. TechRepublic, September 2015, ¡°Executive¡¯s Guide to AI in Business.¡± ¨Ï 2015 CBS Interactive Inc. All rights reserved.
http://www.zdnet.com/article/executives-guide-to-ai-in-business-free-ebook/


7. For more information about Stanford University¡¯s efforts to combine neural networks, computer vision, and natural language processing, visit their website at:
http://cs.stanford.edu/people/karpathy/deepimagesent/


8. For more information about Gartner¡¯s predictions for digital business, visit their website at:
http://www.gartner.com/smarterwithgartner/top-10-strategic-technology-predictions-for-2015-and-beyond/


9. TechRepublic, September 2015, ¡°Executive¡¯s Guide to AI in Business.¡± ¨Ï 2015 CBS Interactive Inc. All rights reserved.
http://www.zdnet.com/article/executives-guide-to-ai-in-business-free-ebook/


10. Business Insider, May 5, 2015, ¡°IBM¡¯s Watson Computer Can Now Do in a Matter of Minutes What It Takes Cancer Doctors Weeks to Perform,¡± by Lauren F. Friedman and Reuters. ¨Ï 2015 Business Insider Inc. All rights reserved.
http://www.businessinsider.com/r-ibms-watson-to-guide-cancer-therapies-at-14-centers-2015-5


11. TechTimes, January 29, 2015, ¡°Bill Gates, Like Stephen Hawking and Elon Musk, Worries About Artificial Intelligence Being a Threat,¡± by Aaron Mamiit. ¨Ï 2015 TechTimes.com. All rights reserved.
http://www.techtimes.com/articles/29436/20150129/bill-gates-like-stephen-hawking-and-elon-musk-worries-about-artificial-intelligence-being-a-threat.htm




The Evolving Role of Artificial Intelligence in Business
 
We discussed the potential impact of artificial intelligence and automation on jobs last month, and we¡¯ve explored the future of driverless cars and trucks in several previous issues. Now, let¡¯s explore how artificial intelligence will impact businesses.


Artificial intelligence (AI) is typically defined as ¡°a field of computer science dedicated to the study of computer software making intelligent decisions, reasoning, and problem solving.¡±


However, that definition leans toward what experts consider ¡°strong AI,¡± which focuses on artificial intelligence systems that are able to perform as flexibly as the human brain. That version of AI is still likely to be at least three decades from becoming a reality.
 
Instead, what is emerging in countless everyday applications today is what is known as ¡°Weak¡± AI. Weak AI functions within a tightly focused area of ability, performing either one or a few simple tasks more efficiently than humans can perform them. Examples include:


- Air-traffic control systems that determine flight plans and choose the optimal landing gates for airplanes.


- Logistics apps that help companies like UPS route their trucks to save time and fuel.


- Loan-processing systems that assess the creditworthiness of mortgage applicants.


- Speech-recognition tools that handle incoming calls and provide automated customer service.


- Digital personal assistants that search multiple data sources and provide answers in plain English, like Apple¡¯s Siri.


In both cases, AI is based on a series of algorithms - a formula or set of rules - that neural networks use to process information and arrive at an answer. While scientists have been working on artificial intelligence for decades, it is only now emerging as an important business tool because of two key developments:


1. Processing power continues to accelerate. As predicted by Moore¡¯s Law, named for Intel cofounder Gordon Moore, the number of transistors per chip has been roughly doubling every year for the past four decades. To keep improving the performance, companies like Nvidia are supplementing the central processing unit (CPU) cores in their chips with graphics processing unit (GPU) cores.1 A CPU consists of a few cores that use serial processing to perform one task at a time, while a GPU consists of thousands of cores that use parallel processing to handle several tasks simultaneously. What this means is that neural networks can master ¡°deep learning¡± applications by analyzing vast amounts of data at the same time.2


2. The amount of data and the amount of data storage - are continuing to expand. According to the research organization SINTEF, ¡°90 percent of all the data in the world has been generated over the past two years.¡±3 And as IDC reports, the amount of data that is created and copied each year doubles in size every two years, so by 2020 there will be 44 zettabytes (44 trillion gigabytes) of data.4 Meanwhile, Kryder¡¯s Law, named for former Seagate chief technology officer Mark Kryder, states that the density of hard disk drives doubles every eighteen months. That means the cost of storage drops by half every eighteen months. This makes it possible to cheaply store all the available data that is needed to ¡°teach¡± neural networks and to aggregate the vast amounts of new data that is being created daily for them to analyze.5


The result of this enormous explosion of data and processing power is already being seen in business applications that can analyze patterns of online purchasing behavior to detect credit-card fraud or determine which ad to show to a particular customer.


As new capabilities are perfected, the possibilities will continue to expand.6 For example, scientists at Google and Stanford University (working separately) have developed systems that combine neural networks, computer vision, and natural language processing.6


The result in both cases is an AI system that can translate an image into text. For example, the systems can correctly identify what is happening in a photograph and provide a description such as ¡°A person riding a motorcycle on a dirt road.¡±


Photographs and videos are part of the unstructured data that had been challenging for AI systems to interpret until now. Consider that this year - largely because of the ubiquity of smartphones - an estimated 1 trillion photographs will be taken, with billions of those photos posted online.


With all of those images available, one application will be the use of AI for security purposes. Recall how the FBI used spectators¡¯ photographs and videos to identify the Boston Marathon bombing suspects. Now multiply the inputs by several trillion and the opportunities for detecting criminal and terrorist activity will expand accordingly.


Robotics is another application that will benefit from advances in AI. For example, the Google and Stanford breakthroughs will allow humanoid robots and driverless vehicles to identify objects in their environment, determine how to interact with them, and describe them to humans.7


Retailers will be able to use the technology (combined with video cameras) to track shoppers in their stores, not just to prevent theft but also to tailor product displays and promotions - such as sending discount coupons to customers¡¯ smartphones.


Looking ahead, we foresee the following developments emerging from this crucial trend:


First, as artificial intelligence becomes increasingly advanced, AI systems will be more widely adopted by businesses.


Improvements in algorithms - and continued progress in computing power and data storage due to Moore¡¯s Law and Kryder¡¯s Law - will transform AI from a tool that leading companies use to gain a competitive advantage today to one that is required simply to stay in business. According to a study by Gartner, artificial intelligence will lower the cost of ownership for business operations by 30 percent by 2019.8 Organizations that fall behind won¡¯t be able to deliver the personalized, low-cost products and services that customers will come to expect. The obvious analogy is to the Internet, which evolved over the past three decades from an additional marketing channel to a powerful disruptive force that reshaped industry value chains and made entirely new business models possible.


Second, artificial intelligence will improve the productivity and accuracy of knowledge workers.


According to TechRepublic, a startup called Kensho ¡°claims to be ¡®the world¡¯s first computational knowledge engine for the financial industry¡¯ - a system that uses massively parallel statistical computing, natural-language inputs, big data, and machine learning to answer complex financial questions posed in plain English.9 It¡¯s a development that could potentially threaten the job security of highly paid ¡®quants¡¯ who are employed to model market dynamics on Wall Street, in London, and hosts of other financial centers.¡± However, it¡¯s more likely that artificial intelligence systems won¡¯t replace highly skilled human workers, at least for the foreseeable future; instead, AI systems will work side-by-side with professionals, providing a powerful tool to boost their effectiveness.


Third, in the healthcare industry, artificial intelligence will revolutionize the prevention and treatment of diseases and other medical conditions.


Consider IBM¡¯s Watson for Oncology system.10 It uses all of the stored wisdom in several medical databases to analyze a patient¡¯s symptoms and then provide a list of suggested treatments (ranked in order of probability of effectiveness) along with a list of references. Also, according to research by Gartner, by 2020, life expectancy in the developed world will increase by an average of six months, thanks to the use of wireless health monitoring technology.


Fourth, while AI will take over many of the tasks that can be automated with algorithms, it won¡¯t be able to mimic the inherently human qualities of human workers.


For instance, Watson for Oncology might even be better than a human doctor at diagnosing cancer in a patient and identifying the best course of treatment, but patients will probably always prefer to hear the news from another human rather than a machine. In environments ranging from the doctor¡¯s office to the sales floor to the hair salon, AI will make processes more efficient, but human workers will still be needed to connect with customers on a personal, emotional level.


Fifth, artificial intelligence will both enable and be enabled by a host of economically transformative technologies.


Just as cars, roads, the petroleum industry, and mass production all evolved together to transform civilization in the early twentieth century, AI is part of a massive set of interdependent industries and technologies that can¡¯t survive without each other. To reach its full potential, it depends on the blossoming of networks, biochips, quantum computing, wearable computers, robotics, speech recognition, machine learning, and the Internet of Things - and all of those technologies depend on AI.


Sixth, the big risk with AI is that, if left unchecked, it could theoretically lead to the end of civilization.


Because strong AI will advance through machine learning in which computers will become progressively smarter as they modify themselves in response to their environment, it¡¯s possible that they could advance beyond our control. The algorithms in today¡¯s AI systems contain rules designed to prevent negative consequences. For example, if the task of a robot vacuum cleaner is defined as ¡°to pick up as much dust as possible in a home,¡± it will continually dump the dust it collects back onto the floor in order to collect it again. Instead, the task must be defined carefully, such as ¡°to keep the floor clean.¡± Similarly, if an AI system were programmed to absorb all of the knowledge on the Internet and ¡°to devise a way to conserve the planet¡¯s resources,¡± it could decide the best solution is to destroy humanity. As far-fetched as that may sound, Stephen Hawking, Bill Gates, and Elon Musk have all expressed concerns about the risks of AI.11 The founding engineer of Skype, Jaan Tallinn, recently created an organization called the Future of Life Institute. With an advisory board that includes many of the world¡¯s leading AI experts and a $10 million donation from Musk, the Institute¡¯s mission is to find a way to prevent the technology from creating a catastrophe.


References
1. For more information about GPU-accelerated computing, visit the Nvidia website at:

http://www.nvidia.com/object/what-is-gpu-computing.html


2. For more information about ¡°deep learning¡± using neural networks, visit the Nvidia website at:

https://developer.nvidia.com/deep-learning


3. For more information about data expansion, visit the SINTEF website at:

http://www.sintef.no/en/corporate-news/big-data--for-better-or-worse/


4. ComputerWeekly, April 9, 2014, ¡°Data Set to Grow 10-Fold by 2020 as Internet of Things Takes Off,¡± by Antony Adshead. ¨Ï 2014 TechTarget, Inc. All rights reserved.

http://www.computerweekly.com/news/2240217788/Data-set-to-grow-10-fold-by-2020-as-internet-of-things-takes-off


5. Scientific American, August 2005, ¡°Kryder¡¯s Law,¡± by Chip Walter. ¨Ï 2005 Scientific American, a division of Nature America, Inc. All rights reserved.

http://www.scientificamerican.com/article/kryders-law/


6. TechRepublic, September 2015, ¡°Executive¡¯s Guide to AI in Business.¡± ¨Ï 2015 CBS Interactive Inc. All rights reserved.

http://www.zdnet.com/article/executives-guide-to-ai-in-business-free-ebook/


7. For more information about Stanford University¡¯s efforts to combine neural networks, computer vision, and natural language processing, visit their website at:

http://cs.stanford.edu/people/karpathy/deepimagesent/


8. For more information about Gartner¡¯s predictions for digital business, visit their website at:

http://www.gartner.com/smarterwithgartner/top-10-strategic-technology-predictions-for-2015-and-beyond/


9. TechRepublic, September 2015, ¡°Executive¡¯s Guide to AI in Business.¡± ¨Ï 2015 CBS Interactive Inc. All rights reserved.

http://www.zdnet.com/article/executives-guide-to-ai-in-business-free-ebook/


10. Business Insider, May 5, 2015, ¡°IBM¡¯s Watson Computer Can Now Do in a Matter of Minutes What It Takes Cancer Doctors Weeks to Perform,¡± by Lauren F. Friedman and Reuters. ¨Ï 2015 Business Insider Inc. All rights reserved.

http://www.businessinsider.com/r-ibms-watson-to-guide-cancer-therapies-at-14-centers-2015-5


11. TechTimes, January 29, 2015, ¡°Bill Gates, Like Stephen Hawking and Elon Musk, Worries About Artificial Intelligence Being a Threat,¡± by Aaron Mamiit. ¨Ï 2015 TechTimes.com. All rights reserved.

http://www.techtimes.com/articles/29436/20150129/bill-gates-like-stephen-hawking-and-elon-musk-worries-about-artificial-intelligence-being-a-threat.htm


ÀÌÀü

¸ñ·Ï