Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
医療 × AI, 量子コンピュータ, ブロックチェーン, ロボット, AR, VR, CRIS...
Search
Sho Okiyama
June 15, 2017
Technology
14
13k
医療 × AI, 量子コンピュータ, ブロックチェーン, ロボット, AR, VR, CRISPR, BCI / Medicine and AI, QC, BC and more
医療 × AI, 量子コンピュータ, ブロックチェーン, and more…
for more info:
http://okiyamasho.com
Sho Okiyama
June 15, 2017
Tweet
Share
More Decks by Sho Okiyama
See All by Sho Okiyama
ChatGPTと医療の現在地 (ver.August-31, 2023)
shookiyama
6
8.6k
エクスポネンシャル・テクノロジーと未来 / exponential technologies and the future
shookiyama
1
11k
イメージで理解する人工知能 〜入門編〜 / understanding AI from scratch
shookiyama
0
9k
Other Decks in Technology
See All in Technology
Turing × atmaCup #18 - 1st Place Solution
hakubishin3
0
490
レンジャーシステムズ | 会社紹介(採用ピッチ)
rssytems
0
200
プロダクト開発を加速させるためのQA文化の築き方 / How to build QA culture to accelerate product development
mii3king
1
270
ハイテク休憩
sat
PRO
2
170
watsonx.ai Dojo #5 ファインチューニングとInstructLAB
oniak3ibm
PRO
0
170
Qiita埋め込み用スライド
naoki_0531
0
5.1k
LINEヤフーのフロントエンド組織・体制の紹介【24年12月】
lycorp_recruit_jp
0
530
KubeCon NA 2024 Recap / Running WebAssembly (Wasm) Workloads Side-by-Side with Container Workloads
z63d
1
250
なぜCodeceptJSを選んだか
goataka
0
160
20241214_WACATE2024冬_テスト設計技法をチョット俯瞰してみよう
kzsuzuki
3
590
re:Invent 2024 Innovation Talks(NET201)で語られた大切なこと
shotashiratori
0
320
WACATE2024冬セッション資料(ユーザビリティ)
scarletplover
0
210
Featured
See All Featured
Practical Tips for Bootstrapping Information Extraction Pipelines
honnibal
PRO
10
810
Chrome DevTools: State of the Union 2024 - Debugging React & Beyond
addyosmani
2
170
How GitHub (no longer) Works
holman
311
140k
Making Projects Easy
brettharned
116
5.9k
The Myth of the Modular Monolith - Day 2 Keynote - Rails World 2024
eileencodes
17
2.3k
Thoughts on Productivity
jonyablonski
67
4.4k
Java REST API Framework Comparison - PWX 2021
mraible
28
8.3k
Principles of Awesome APIs and How to Build Them.
keavy
126
17k
Exploring the Power of Turbo Streams & Action Cable | RailsConf2023
kevinliebholz
28
4.4k
Building Better People: How to give real-time feedback that sticks.
wjessup
365
19k
The Cost Of JavaScript in 2023
addyosmani
45
7k
I Don’t Have Time: Getting Over the Fear to Launch Your Podcast
jcasabona
29
2k
Transcript
ҩྍ ʷ AI, ྔࢠίϯϐϡʔλ, ϒϩοΫνΣʔϯ, and more… 2017.06.15 ܦςΫϊϩδʔNEXT 2017
ԭࢁ ᠳ
“21ੈلͷऴΘΓ·ͰʹશͯͷපؾΛ ࣏ྍ͠ɺ༧͠ɺ͋ΒΏΔපؾʹ ରԠͰ͖ΔΑ͏ʹ͍ͨ͠” –– ϚʔΫɾβοΧʔόʔά CC-BY SA, Alessio Jacona, flickr
“2030ɺ͋Δ͍͘ͱ 2030ͷऴΘΓ·Ͱʹɺ ਓྨશͯͷපؾΛ Γӽ͑ΒΕΔͩΖ͏” –– ϨΠɾΧʔπϫΠϧ TED, https://www.youtube.com/watch?v=PVXQUItNEDQ
0% 20% 40% 60% 80% 100% ༧ఆ ࣮ࡍ ςΫϊϩδʔͷਐาࢦؔత ώτήϊϜܭը
ਐḿ ʢ60ԯUSD/15ϲʣ 1997 ήϊϜͷղಡୡ1% ଟ͕͘ܭըதࢭΛओு ͔͠͠ʮ 1% ऴΘͬͨͷͳΒɺ͏ऴΘͬͨʹ͍͠ʯ 1%
0% 20% 40% 60% 80% 100% ༧ఆ ࣮ࡍ ςΫϊϩδʔͷਐาࢦؔత ώτήϊϜܭը
ਐḿ ʢ60ԯUSD/15ϲʣ
0% 20% 40% 60% 80% 100% ༧ఆ ࣮ࡍ ςΫϊϩδʔͷਐาࢦؔత ώτήϊϜܭը
ਐḿ ʢ60ԯUSD/15ϲʣ ʮ୭ࣗ͠ͷ૾ྗͷݶք͕ɺੈքͷݶքͩͱޡղ͢Δʯ ʢγϣʔϖϯϋΞʔʣ
ਓೳ (AI)
ͳ͍ͥ·AIͳͷ͔ ΫϥυԽʹΑΔେྔͷσʔλ େྔͷσʔλ GPUʹΑΔฒྻॲཧɺܭࢉͷ্ ϋʔυΣΞͷਐԽ PepperɺΞϧϑΝޟ ੈ͔ؒΒͷظײ Ωϟονʔͳࣄྫ
AIͷϨΠϠʔ ਓೳͱݺΕͳ͘ͳͬͨͷ ਓೳ ػցֶश σΟʔϓϥʔχϯά ϧʔϧϕʔευ l AlphaGo l ը૾ೝࣝ
l AI in the 1980’s l ࣗಈυΞ l ϧϯό l ʮAͳΒBʯ l MS OfficeͷΠϧΧ ൚༻ਓೳʁ ݱঢ়ͷ Ԇઢ্ʹͳ͍
σΟʔϓϥʔχϯά ͜Ε·Ͱ ཱ͕ࣖͬͯͯ… ͕2ͭ͋ͬͯ… ඓࡾ֯Ͱ… ͦͷԼʹޱ͕… ͯ͢ਓ͕ڭ͍͑ͯͨ - ώή͋Δͷʁͳ͍ͷʁ -
ޱԿݸʁ - ͋Εʁ͜Εʁ ԿΛͲ͜·Ͱڭ͑Δ͖͔ෆ໌ ݱࡏ ػց͕উखʹڞ௨Λֶश - ࣅͨࣸਅ͝ͱʹ͚ͱ͍ͯʂ - ͋ɺ͜ͷࢁ͕ʮωίʯͶɻ αϯϓϧ͕૿͑Εݸମࠩٵऩɺ ฏۉ͢ΔͱʮೣΒ͠͞ʯ͕Δ
ʦᗡྨʧ τΧήͱͶɺϫχͱ ΧϝͱΠϞϦͱ… αϯϓϧ͑͞༩͑Εɺ ʮᗡྨͱ͍͏֓೦ʯͷநग़ AIͤͰOK ʮಛྔʯͱ ʦԁʧ ͋Δ͔Βͷڑ͕ ͍͠ͷू߹
ఆٛͰ͖Δͷѻ͑Δ ͜Ε·Ͱ σΟʔϓϥʔχϯά ʦᗡྨʧ ͑ʔɺมԹಈͰ ͕4ຊ͋ͬͯ… ੑ࣭ྻڍͷΈͰѻ͑ͳ͍ ఆٛͮ͠Β͍நతͳ֓೦ ੑ࣭ͷ··ʢ=ಛྔʣѻ͑Δ
ಛྔͷԋࢉ ֆͷ୯७ͳ߹Ͱͳ͘ɺ “ΰοϗΒ͠͞” ͱ͍͏ ಛྔͷऔΓѻ͍͕Մೳʹ + https://deepart.io/ ݩֆ ը෩
ҩྍ ʷ AI ELIZA (1966) MYCIN (1975) ͲͪΒʮύοͱݟʯ͍͕͢͝ɺ ҩྍݱʹೖΓࠐΊͳ͍··AIౙͷ࣌… ײછͷ࣏ྍํΛఏࣔɻҰൠՊҩΛ্ճΔਖ਼֬ੑ
ʢϧʔϧϕʔευͳٛͷ ”AI” ʣ ΧϯηϦϯάࣜͷରܕϓϩάϥϜ ʢੈքॳͷνϟοτϘοτʣ
ҩྍ ʷ AI ELIZA (1966) MYCIN (1975) ͲͪΒʮύοͱݟʯ͍͕͢͝ɺ ҩྍݱʹೖΓࠐΊͳ͍··AIౙͷ࣌… ײછͷ࣏ྍํΛఏࣔɻҰൠՊҩΛ্ճΔਖ਼֬ੑ
ʢϧʔϧϕʔευͳٛͷ ”AI” ʣ ΧϯηϦϯάࣜͷରܕϓϩάϥϜ ʢੈքॳͷνϟοτϘοτʣ Siri (2011) Watson (2011) ͦͯ͠σΟʔϓϥʔχϯάͷ࣌
- ݟಀ͠ɿ0%ʢਓؒ7%ʣ - ޡɿ30%ݮʢਓؒൺʣ - ॲཧɿ5ສഒʢਓؒൺʣ ΦʔετϥϦΞͰ࣮ূ࣮ݧதʢ40ҩྍػؔʣ ϨϯτήϯCTͷը૾அ ਓؒʹΑΔݟಀ͠ޡஅݱঢ় 3ສ݅/dayʢถࠃʣ
ʮAI͕ݟಀͨ͠ΒͲ͏͢Δͷʁʯ enlitic http://www.enlitic.com/
Atomwise - ಛఆͷλϯύΫ࣭ʹ࡞༻͍͢͠ࢠߏΛ εΫϦʔχϯά - ඞཁظؒΛඦͷҰʹॖ - ΤϘϥΠϧεͷ࣏ྍༀީิΛൃݟ ৽ༀ୳ࡧΛޮԽ ೳಈతͳࢠݕࡧ͕ʑՃ
ༀྖҬͷϒϨΠΫεϧʔ http://www.atomwise.com/
ͦͷଞ ͷऔΓΈ Google “Project Baseline” DeepMind Health + NHS Stanfordେ
ൽෘ͕Μͷը૾அ GoogleࡿԼDeepMind + ӳࠃࠃຽอݥαʔϏε ؟Պ࣬ױʢஅ+࣏ྍʣ͔Βணख ରऀ1ສਓɺ1ԯυϧ/4ؒͷϓϩδΣΫτ ήϊϜɺ݂ӷɺɺଣӷɺྦӷ etc. ݱ൛ϑϥϛϯΨϜݚڀ ൽෘՊઐҩΛ্ճΔ ਓ͕ؒண͠ͳ͍ɺ͕Μࡉ๔पғͷࡉ๔ͷ༷ࢠ͔Βஅ
σΟʔϓϥʔχϯάͷ՝ 1 Πν͔Βֶश͢Δʹେྔͷదͳαϯϓϧ͕ඞཁ σʔλͷ࣭ͱྔ AI͕ೝࣝͨ͠ʮμϯϕϧʯ https://research.googleblog.com/2015/06/inceptionism-going-deeper-into-neural.html
σΟʔϓϥʔχϯάͷ՝ 2 ͲΕωίͳͷʹɾɾɾ ֶशσʔλ֎ͷԠ༻͕ར͔ͳ͍
σΟʔϓϥʔχϯάͷ՝ 3 AIͷཧղ͕ਓؒʹײతͰͳ͍ɻ ࠜڌෆ໌ͷϒϥοΫϘοΫε ײͷҼղ ʢਓؒྲྀʣ ߹ཧత͕ͩɺ”X” ʹ૬͢Δײ ਓؒʹΠϝʔδͮ͠Β͍ ײ
ت ײ ”X” ײͷҼղ ʢAIྲྀʣ ʮ4छྨෆཁͰʁʯ ʮتͱָॏෳ͍ͯ͠·͢ʯ ʮѩ = ϚΠφε تʯ ت ָ ౖ ѩ ײ
՝͋ΕͲʑղܾ ػցֶशɺΑΓਓؒΒ͘͠ ಄ͷதΛը૾Ͱදݱ - GAN (Generative Adversarial Network)ɿ ֶश༰Λը૾Խͯ͠ग़ྗɻਓؒʹͱͬͯײతͰɺ ·ͨग़ྗը૾ΛֶͬͨࣗࣗशՄೳ
- సҠֶशɿΞϝϑτͰֶΜͩܦݧ͕ϥάϏʔʹ׆͔ͤΔ - one-shot/zero-shot learningɿॳΊͯϖϦΧϯΛݟͯɺ ͦΕ͕ௗͷҰछͩͱ͔Δ
ҩྍʹ͓͚ΔAIͷՄೳੑ 1 1ର1ͷؔ ʮif ڳ͕௧͍ then ৺ଁපͩʯ Nର1 ͷஅ 2ର1ͷؔ
ʮif ͕ߴ͍ΘΓʹ຺͕ૣ͘ͳͬͯͳ͍ then ϨδΦωϥײછͩʯ → ΄΅ਓؒͷݶք Nର1ͷؔ ʮif ߴྸͰOܕRh+ͷΘΓʹ݂͕͋Δ͕֏ͳ͍ then …ʯ → ม͕ଟ͗ͯ͢ਓؒʹॲཧෆೳ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔ ˔
ҩྍʹ͓͚ΔAIͷՄೳੑ 2 ใࠂ͔Βྟচͷ͕࣌ؒࠩθϩʹ ױऀͷಛϕΫτϧݕࡧ + จཏ + ࣗಈ༁ ඇӳޠݍͷҩֶࢽ͔ΒྨࣅྫΛݟ͚ͭग़͢ CC-BY
SA, Crates, Wikimedia Commons
ͦͷଞ ੑͷߴ͍ྖҬ ҩࢣΞϥʔτ ݕ࣏ࠪྍͷൈ͚࿙ΕɺༀࡎҿΈ߹ΘͤνΣοΫɺetc. පؾͷൃεΫϦʔχϯά ΣΞϥϒϧ͔Β৺ͷҟৗݕɻѱԽલͷडఏҊ ೖӃۀޮԽ ױऀݕࠪͷॱ൪ɺࣄނͷະવࢭɺೖӃ෦ͷޮՌతׂৼ ը૾அɺༀ ઌड़ࣄྫհͷ௨ΓɺAIΞϧΰϦζϜͱ૬ੑΑ͍
ϩϘοτ
ϩϘοτͷͨΒ͢Ձ ࣗಈԽ ӡಈɾײ֮ೳྗͷ֦ு - ޮԽͱ࠶ݱੑ্ - ਓؒ༗ࣄͷରԠʹ - ਫ਼ͱࣗ༝ͷߴ͍ಈ͖ -
ޒײͷೳྗ૿ڧ - ൪Ҏ߱ͷײ֮֫ಘ
- ମྗ૿ڧɿ࠲ͬͯखज़Ͱ͖Δ + ΞʔϜϨετ - ࢹྗ૿ڧɿ3Dࢹڸ + ζʔϜࢹ 15ഒ -
ख͕3ຊʹɿ3ຊͷΞʔϜΛ1ਓͰૢΕΔ - ؔઅػೳ૿ڧɿؔઅՄಈҬ360 - ਆܦɺ ےػೳ૿ڧɿࣗಈखͿΕิਖ਼ - খػೳ૿ڧɿϞʔγϣϯεέʔϧʢ1:5ʣ μɾϰΟϯν Intuitive Surgical https://www.intuitivesurgical.com/ ಓ۩Λ͑ͨɺਓؒͷೳྗ૿ڧσόΠε
- Johns Hopkinsେͱڞಉ։ൃ - 2016 ಲͷΛࣗಈ๓߹ ਓؒΑΓਖ਼֬ - ࢹ֮ɿ֎ઢ +
ۭؒߏͷѲ - ৮֮ɿͷײ৮ɺ๓߹ࢳͷక·Γ۩߹ - ๓߹Ͱେͳ͜ͱɿ ѱ͍ͻͱΛݮΒ͢ ʼ ྑ͍ͻͱΛ૿͢ STAR (Smart Tissue Autonomous Robot) Intuitive Surgicalʢঝલʣ https://www.intuitivesurgical.com/
ͦͷଞ ͷऔΓΈ Verb Surgical Virtual Incision खͷͻΒαΠζͷμɾϰΟϯνܕϩϘοτ ෲ෦ʹ։͚͔ͨ݀Βૠೖ͠ɺମ͔Βखज़Λߦ͏ Googleͱ JˍJ
͕ࣾڞಉͰઃཱɺखज़ΞγετϩϘοτ τοϓ5%֎Պҩͷٕज़Λ୭Ͱ࠶ݱՄೳʹ
ҩྍʹ͓͚ΔϩϘοτͷՄೳੑ - ҩྍऀଆ ਓతϦιʔεͷସ ઐٕज़ͷίϞσΟςΟԽʢӡಈೳྗ֦ுʣ ࢦઌ͑ͣɺ๓߹ۉҰͰɺ ࢹڸนʹͨΒͣɺࢭ݂࣮֬ʹ ࠾݂ɺణɺ ؆қखज़ͷࣗಈԽʢϨʔγοΫɺൽෘजᙾআڈʣ ࣄલใྔͷ૿Ճʢ֮ೳྗ֦ுʣ
ηϯαʔVR, ARٕज़ͱͷ౷߹Ͱखज़෦ҐͷԹɺ ࣓ྗɺ࣭ͷ࣭ײີ͕ʮݟ͑ΔʯΑ͏ʹ
ҩྍʹ͓͚ΔϩϘοτͷՄೳੑ - ױऀଆ ϦϋϏϦςʔγϣϯ ϩϘοτίϯύχΦϯ PepperϖοτܕϩϘοτ ࣾձతݽཱ৺ͷ݈߁ͱ૬ؔɻ ϞχλϦϯάɺίϛϡχέʔγϣϯσόΠεΛ݉ͶΔ ϦϋϏϦͷ্࣭ɺԼͨ͠ӡಈػೳͷସ
VR, AR
VR, ARͱ ݱ࣮ੈք Ծݱ࣮ ʢVirtual Realityʣ ֦ுݱ࣮ ʢAugmented Realityʣ VRɿ100%
ԾۭؒɻͦͷͰ͏ͷ͕લఏ ARɿݱ࣮ۭؒݟ͑Δɻಈ͖ͳ͕ΒମݧՄೳ
MRͱ MR (mixed reality) ݱ࣮ͷ֦ுͱ͍͏ཱͷ AR ʹରͯ͠ɺ ݱ࣮ɺԾੈքͷڥքͷͳ͞Λڧௐͨ͠ͷ͕ MRɻ Έಉ͡ɺੈք؍ͷҧ͍
ʢຊࢿྉͰ “AR” ʹ౷Ұʣ
େऺ͚ϞσϧήʔϜ͔Β ఱಊɺ1995 όʔνϟϧϘʔΠ (VR) ϙέϞϯGO (AR) NianticʢGoogleΑΓಠཱʣ 2015
2016/4~ɿϧʔϜεέʔϧػೳͰ෦ΛVRۭؒʹ දతͳVRϋʔυ Oculus Rift HTC Vive PlayStation VR Google Cardboard
2014/6~ɿຊମεϚϗΛར༻ 2016/10~ɿຊޠɺՁ֨ɺPS4Ͱ࡞ಈ 2016/3~ɿFacebookʹΑΔങऩ (2014, $2B)
දతͳARϋʔυ Google Glass 2013/2~2015/1ɿ ϓϥΠόγʔݒ೦͔ΒҰൠൢചதࢭʹ Microsoft HoloLens 2016/3~ɿ։ൃऀ͚ϞσϧͷΈൢച Magic Leapʢະൃച
ະެදʣ ϓϩμΫτશൿಗͷ··اۀՁ $4.5B
ARͷ֩ͱͳΔٕज़ μϒϧΧϝϥʢԞߦ͖ೝࣝʣɺGPSɺ࣓ؾηϯαʔɺ ՃηϯαʔɺδϟΠϩηϯαʔ ݱ࣮ۭؒͷѲ AR vs ϦΞϧσʔλͷಉҰ࠲ඪॲཧ ϢʔβʔߦಈʹΑΔΠϯϓοτ ࢹઢํ͕มΘͬͯɺ ARσʔλݱ্࣮ۭؒͰͦͷ··
Ϣʔβ͕ͱͬͨߦಈΛೝࣝ͠ɺ ARۭؒʹϑΟʔυόοΫө
CT, MRIը૾ΛݩʹױऀͷମΛVRͰ࠶ߏ खज़ͷγϛϡϨʔγϣϯɺҩֶੜͷڭҭʹ׆༻ Surgical Navigation Advanced Platform Surgical Theater http://www.surgicaltheater.net/
Surgical Theater, https://www.youtube.com/watch?v=8DnEadLgWCE
VRΛ༻͍ͨखज़γϛϡϨʔγϣϯ ज़ऀʹ৮֮ϑΟʔυόοΫ͕ฦΔ खज़܇࿅γϛϡϨʔλʔ Immersive Touch http://immersivetouch.com/ ImmersiveTouch, https://vimeo.com/197613091
ถࠃͰ܉ਓͷPTSDൃ͕ʹ ࿐ྍ๏ͷҰछͱͯ͢͠ͰʹΤϏσϯε͋Γ Brave Mind Virtually Better http://www.virtuallybetter.com/ http://www.virtuallybetter.com/
VR, ARͷͨΒ͢Ձ ૣૹΓʹΑΔ࣌ؒॖɺ࣌ؒٯҠಈʹΑΔΓ͠ VRੑɿ࣌ؒͷૢ࡞ੑ VRੑɿۭؒͷૢ࡞ੑ ARੑɿޒײʢϞμϦςΟʣͷ֦ு ߦ͚ͳ͍ͱ͜Ζߦ͚Δɺ͙͢ʹߦ͚Δɺ ରͷ֦େ&ॖখ ηϯαʔIoTͱͷ࿈ܞͰ ݟ͑ͳ͍ͷ͕ʮݟ͑ΔʯԽ
ҩྍʹ͓͚ΔVR, ARͷՄೳੑ γϛϡϨʔγϣϯɺ܇࿅ ࢹྗ֦ு खज़ࣨͰମߏɺजᙾपғͷಈ຺ΛՄࢹԽ ࣦഊ͕ڐ͞ΕΔڥԼͷ܇࿅ ױऀͷϏδϡΞϥΠθʔγϣϯ ݴ༿Ͱͷઆ໌͔Βɺମݧ͢Δઆ໌
ϒϩοΫνΣʔϯ
ϒϩοΫνΣʔϯͱ σδλϧใͷਅِΛอূ͢Δٕज़͕ϒϩοΫνΣʔϯ ΠϯλʔωοτใΛѻ͏ ใίϐʔɺվ᜵Ͱ͖ͯ͠·͏ ଟ͘ͷͷ͕ใԽ͞Εɺ ίϐʔվ᜵͞ΕΔͱࠔΔͷ͔Γʹ
ΈʢϏοτίΠϯΛྫʹʣ - औҾΛ·ͱΊͯϒϩοΫͱͯ͠ѻ͏ - ϒϩοΫʹաڈͷऔҾ݁Ռ͕ه͞Ε͍ͯΔ - 10͝ͱʹܨ͕ͬͯ৳ͼ͍ͯ͘ A → B
A → B → C A → B → C → D ɾ ɾ ɾ ϒϩοΫ + νΣʔϯ = ϒϩοΫνΣʔϯ
ΈʢϏοτίΠϯΛྫʹʣ A → B A → B → C A
→ B → C → D ɾ ɾ ɾ ʷ Z (10min) ෆਖ਼ॻ͖͑10Ҏʹࡁ·ͤͳ͍ͱɺ
ΈʢϏοτίΠϯΛྫʹʣ ෆਖ਼ॻ͖͑10Ҏʹࡁ·ͤͳ͍ͱɺ ϒϩοΫ͕࣍ʑͱ৳ͼͯΠλνͬ͜͝ʹ A → B A → B →
C A → B → C → D A → B → C → D → E ʷ Z
͔͠શһ͕ಉ͡νΣʔϯΛ͍ͬͯΔ ʮෆਖ਼͕Ͱ͖ͳ͍औҾهாʯ A→B ɾ ɾ ɾ ɾ ɾ ɾ A→B
ɾ ɾ ɾ ɾ ɾ ɾ A→B ɾ ɾ ɾ ɾ ɾ ɾ A→B ɾ ɾ ɾ ɾ ɾ ɾ
ใͷࢄཧ ͜Ε·Ͱ ϒϩοΫνΣʔϯ - ࠃɺاۀ͕ใΛཧ - ϢʔβʔใΛ ΘͤͯΒ͏ཱ - தԝෆࡏ
- Ϣʔβʔશһ͕ର
ϒϩοΫνΣʔϯͷͨΒ͢Ձ ใͷσδλϧԽ - ෳɺվมͷڪΕ͕ݮΓɺσδλϧԽ͕ଅਐ - ྫɿॴखଓ͖ɺূ໌ɺࢿ֨ɺܖॻ ख͞ͳ͍͕ࣾձͷΠϯϑϥͱͳΔٕज़ தԝཧऀͷෆࡏ ݖྗମ͔Βͷґଘ -
ࠃɺେاۀʹཔΒͳ͍ࣾձ ݖྗମʹཔΔϦεΫɿಠࡋԽɺτοϓͱͷڞΕ - ϦεΫྫɿ՟ฎ → ͕ΕͯϋΠύʔΠϯϑϨ - ϦεΫྫɿۚ → ཧෆඋʹΑΓೲه͕ফࣦ
ҩྍʹ͓͚ΔϒϩοΫνΣʔϯͷՄೳੑ ҩྍهͷݸਓཧ ҩༀͷτϨʔαϏϦςΟ ݙ݂ࡎɺҨࢠΈ͑ࡎ ҩྍऀͷशख़ධՁɺ࿑ཧ ୭͕Կͷखज़ΛԿ͔݅ͬͨɺ࣌ؒ֎࿑ಇ࣌ؒͷѲ ެ։ݖݶͷױऀอ༗ɺੜ֔ΞΫηεϩά
ͷऔΓΈ DeepMind Health MIT MedRec ҩྍใͷެ։ൣғΛױऀ͕ཧ ϒϩοΫνΣʔϯͰใͷਖ਼ਅੑΛ୲อ ӳࠃࠃຽอݥαʔϏε (NHS) ͕ױऀσʔλΛఏڙ
ࣗͷใΛͲ͜Ͱ୭͕༻͍͔ͨɺױऀ͕ࣗ֬ೝՄೳʹ
ྔࢠίϯϐϡʔλ
ྔࢠίϯϐϡʔλͱ ʮैདྷͷίϯϐϡʔλͷ1ԯഒ͍ʯ – Google ैདྷͷίϯϐϡʔλͰ ղ͖ಘͳ͍͕ղ͚Δ ʮ21ੈلޙͰ…ʯ ۙ5Ͱେ͖ͳਐలɻݚڀ͔Β༻ԽϑΣʔζ
ίϯϐϡʔλͷछྨ ݹయత ίϯϐϡʔλ ྔࢠίϯϐϡʔλ (QC: Quantum Computer) ྔࢠήʔτࣜQC ྔࢠΞχʔϦϯάࣜQC ͦͷଞͷ
“ίϯϐϡʔλ” DNAίϯϐϡʔλ ೪ەίϯϐϡʔλ εύήςΟίϯϐϡʔλ etc. ݱࡏͷίϯϐϡʔλ σδλϧίϯϐϡʔλ Ξφϩάίϯϐϡʔλ
ྔࢠίϯϐϡʔλͷछྨ ྔࢠήʔτࣜ ྔࢠΞχʔϦϯάࣜ - ࠓͷίϯϐϡʔλͱࠜຊతʹ֓೦͕ҟͳΔɺཧγϛϡϨʔλ - ༻్ݶΒΕΔ͕͜ͷܕͷΈʹղ͚Δ͕ଘࡏʢ࠷దԽʣ - D-WaveࣾΑΓ2011ʹൃച։࢝ -
ࠓͷίϯϐϡʔλͷສഒૣ্͍Ґޓʢཧ্ʣ - ϋʔυখϏοτͷϓϩτλΠϓͷΈଘࡏ - ιϑτ·ͩ·ͩݚڀஈ֊ͰΞϧΰϦζϜछྨগͳ͍
ྔࢠͷੑ࣭ ཻࢠͰ͋ΓɺͰ͋Δ ཻࢠɿͲͪΒ͔ҰํΛ௨Δ ɿಉ࣌ʹ྆ํΛ௨Δ
ඇྔࢠͱྔࢠ 1 0 1 0 τϥϯδελʢ=εΠονʣ ྔࢠήʔτࣜ ྔࢠίϯϐϡʔλ Ұʹ 24
= 16௨ΓΛಉ࣌ॲཧ ʢNྔࢠϏοτͳΒ 2N ௨Γʣ 1 1 1 0 1 0 1 0 ैདྷͷίϯϐϡʔλ - 4ϏοτͰ “1101” Λද͍ͯ͠Δ - ҰʹҰ௨Γͣͭॲཧ ྔࢠͷฒߦॲཧੑ ʢ=͍ʣ Λར༻ͨ͠ͷ͕ྔࢠήʔτࣜQC
ྔࢠήʔτࣜ ྔࢠΞχʔϦϯάࣜ ͏Ұͭͷํࣜ
͍͍ͯΔ ࠷దԽ ༩͑ΒΕͨ݅ͰඪΛ࠷খԽ͢Δ ८ճηʔϧεϚϯ NΫΠʔϯ ˒ ˒ ˒ ˒ ˒
˒ ˒ ˒ ˔ ˔ ˔ ˔ ˔ ˔ ˔ Ұචॻ͖ͷ૯ڑΛ࠷খԽ͢Δ NʷN֨ࢠʹνΣεͷΫΠʔϯΛNݸฒɺ ܦ࿏্ͷॏෳΛ࠷খԽ͢Δʢແ͘͢ʣ
ͷΞϓϩʔνʢྔࢠΞχʔϦϯάࣜ ʣ ͖Ε͍ʹຒΊͳ͍͞ ʮ༨ന͕θϩʹۙͮ͘Α͏ɺஔΛ࠷దԽʯ ༨ന͕ଟ͍ͱෆ҆ఆʹͳΔྔࢠϞσϧΛΉ ༨ന = 4 ༨ന =
4 ༨ന = 0 ͳ͓ ճྫ ݅ ํ
ʮ༨ന͕ଟ͍ͱෆ҆ఆʹͳΔྔࢠϞσϧʯͷྫ ֤Ϛεʹʮࣗͷ্Լࠨӈͷ༨നʯΛ༩͑Δ ͦͷ૯Λ࠷খԽ͢Δ͜ͱΛߟ͑Δ 1 2 1 1 1 1 2
2 2 2 2 3 2 1 2 1 1 1 =25 0 1 1 1 1 1 2 2 0 1 2 1 0 0 0 1 =14 ్த·Ͱྑ͍͕ ख٧·Γ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 =0 0 0 0 0 1 0 0 1 1 2 1 0 2 1 1 1 =11 ࠷దղ ʢຊ͜͏ߦ͖͍ͨʣ
ߴ͍ͱ͜Ζ͔Β͍ͱ͜Ζ ← ϒϩοΫͷஔύλʔϯ → ྔࢠ ࠷Լ͕ ༨നθϩʹ૬ A B C
D E F G H I J K L M N O P ߦ͖ࢭ·ΓͰɺ 2าԼ͕ͬͯ3าਐΉͷ͕ ैདྷ ʢඇྔࢠʣͷΓํ ଟ ༨ ന ͷ ૯ গ ࠷దղ
ྔࢠΤωϧΪʔతʹ҆ఆ͢Δঢ়ଶΛΉ ෆ ҆ ఆ Τ ω ϧ Ϊ ồ ҆
ఆ ← ϒϩοΫͷஔύλʔϯ → ʢϏοτ͕ࣔ͢ύλʔϯʣ ̍ ̍ ̍ ̍ ̍ ̍ ̍ ̌ ̍ ̍ ̌ ̍ ̍ ̍ ̌ ̌ ̍ ̌ ̍ ̍ ̍ ̌ ̍ ̌ ̍ ̌ ̌ ̍ ̍ ̌ ̌ ̌ ̌ ̍ ̍ ̍ ̌ ̍ ̍ ̌ ̌ ̍ ̌ ̍ ̌ ̍ ̌ ̌ ̌ ̌ ̍ ̍ ̌ ̌ ̍ ̌ ̌ ̌ ̌ ̍ ̌ ̌ ̌ ̌ ྔࢠ ҆ఆՕॴ͔Β ࣗવʹఆ·Γɺ ྔࢠ͕ͦΕΛݟ͚ͭΔ ଟ ༨ ന ͷ ૯ গ ྔࢠͷτϯωϧޮՌɿ ࠷దղʹࢸΔ·Ͱɺ ߦ͖ࢭ·ΓΛ͢Γൈ͚ͯ͠·͏
ͦͦ ͍Λղ͘ ͱʁ ͍Λ݅ͷܗʹॻ͖Լ͢ ΞϧΰϦζϜΛߟ͑Δ ݅ʹΞϧΰϦζϜΛద༻͢Δͱɺ ղ͕ಘΒΕΔ 1 2 3
͍ ैདྷͷΞϓϩʔν ॴ࣋ۚ800ԁͰɺ100ԁͷϦϯΰͱ50ԁͷϛΧϯΛ ߹Θͤͯ10ݸങ͍͍ͨɻͲ͏͢Δʁ “x + y = 10” Λ50ഒͯ͠ɺલऀ͔ΒҾ͘
͋ͱίϯϐϡʔλͤͰOK͕ͩɺ ΞϧΰϦζϜਓ͕ؒݟ͚ͭͳ͍ͱղ͚ͳ͍ 100x + 50y = 800ɺx + y = 10 ݅ ΞϧΰϦζϜ
͍ ྔࢠΞχʔϦϯάࣜ QC Ͱ͜͏ͳΔ ॴ࣋ۚ800ԁͰɺ100ԁͷϦϯΰͱ50ԁͷϛΧϯΛ ߹Θͤͯ10ݸങ͍͍ͨɻͲ͏͢Δʁ ʢෆཁʣ 100x + 50y
= 800ɺx + y = 10 ݅ ΞϧΰϦζϜ ຬ͖ͨ͢ʪ݅ʫ͑͞͏·͘ೖྗ͢Εɺ ͋ͱྔࢠίϯϐϡʔλ͕ࣗવʹղΛݟ͚ͭͯ͘ΕΔ
ྔࢠΞχʔϦϯάࣜ QC ͷߟ͑ํ ͜Ε·Ͱ ྔࢠΞχʔϦϯά - ΞϧΰϦζϜ - ղͷॴ͕ϐϯϙΠϯτ Ͱ͔Δʢղͷਤʣ
- ʮղ (x, y) = (6, 4) Ͱ͢ʯ ΞϧΰϦζϜʢਤʣศརɻ͔͠͠ਤ͕ͳ͍ɺ ղͷํ֯Λࢦ͢ίϯύε͕͋Ε͑ʹḷΓண͚Δ - ࣮ݧɺγϛϡϨʔγϣϯ - ݱࡏ͔Βͷղͷํ͕ ͔Δʢղͷίϯύεʣ - ʮࠓ (x, y) = (10, 12) ʹ͍Δʁ ߦ͖ա͔͗ͩΒͬͯͶʯ
ղ͚Δͷൣғ ͜ͷੈͷ͋ΒΏΔ͍ ݅Λॻ͖ԼͤΔ ΞϧΰϦζϜൃݟࡁΈͷ ΞϧΰϦζϜൃݟࡁΈ ͔ͭɺ ৗࣝత࣌ؒʹղ͖ऴΘΔ ݹయతίϯϐϡʔλͰ ࠓ·Ͱղ͚ͨ ݹయతίϯϐϡʔλͷൃలͰ
কདྷղ͚Δ͔͠Εͳ͍ ྔࢠίϯϐϡʔλͰ ղ͚Δ͔͠Εͳ͍ ݅ʹॻ͖Լͤͳ͍ ʢྫɿਓੜͷతԿ͔ʁʣ
- ॳ߸ػ2011ൃചɿ128ྔࢠϏοτ - D-Wave 2Xɿ1000ྔࢠϏοτ - 2017ɿ2000ྔࢠϏοτͷ৽४උத - Google/NASA, ϩοΩʔυϚʔςΟϯ͕ࣾߪೖ
- ҩྍͷʹݟ͑ΔԠ༻͜Ε͔Β D-Wave 2XʢྔࢠΞχʔϦϯάࣜʣ D-Wave https://www.dwavesys.com/
ͦͷଞ ͷऔΓΈ IBM Quantum Experience GoogleྔࢠਓೳݚڀॴʢQuAILʣ IARPAʢೳߴݚڀܭըہɺถʣ 2013ઃཱɻ2022·ͰʹϋΠϒϦουͷ ྔࢠίϯϐϡʔλΛ༻Խ͢Δ༧ఆ ྔࢠήʔτࣜͰ5ྔࢠϏοτɻΫϥυͰૢ࡞Մೳ
ΑΓߴػೳͷྔࢠίϯϐϡʔλ “Q” Λ४උத ྔࢠΞχʔϦϯάࣜ࡞தɻ100ྔࢠϏοτ͕ͩ Ϗοτಉ࢜ͷଓ͕ཏతͰߴੑೳ
ҩྍʹ͓͚ΔྔࢠίϯϐϡʔλͷՄೳੑ ༀྖҬ ์ࣹઢ࣏ྍͷরࣹεςϥςδʔཱҊ ࠷దԽͱͯ͠ղੳՄೳ - λϯύΫ࣭ͷߏܾఆɻޮՌతͳΞϛϊࢎͷ ંΓͨͨΈߏΛಉఆ - ಛఆͷड༰ମʹ࡞༻͢ΔࢠͷεΫϦʔχϯά ػցֶशͷՃ
AIͱͷ૬ޮՌͰը૾அɺݕࠪͷޮԽɺetc.
CRISPR
CRISPRʢΫϦεύʔʣͱ ҨࢠΛࣗࡏʹॻ͖͑Δٕज़ 1970’s Ҩࢠ 1990’s ήϊϜฤू ୈ1ੈ 2000’s ήϊϜฤू ୈ2ੈ
2012 CRISPR ੜ B.C. (Before CRISPR) ݱ
CRISPRͷΈ ݩʑࡉە͕ΛकΔ໔ӸγεςϜ ࡉە Πϧε ࡉە DNA ఢͷDNAΛه͔ͯ࣍͠Βૉૣ͘ରԠ
CRISPRͷΈ CRISPR CRISPR Cas9ʢ߬ૉʣ ࡉە͕ΠϧεΛऔΓࠐΉɺ DNAͷΧοτ&ϖʔετख๏ ͜ΕΛԠ༻ٕͨ͠ज़͕ “CRISPR-Cas9” ௨শCRISPR ҹͱͳΔҨࢠྻͷ͜ͱ
(Clustered, Regularly Interspaced, Short Palindromic Repeats)
ओͳԠ༻ྖҬ ੜ͕ؔΘΔ͕ͯ͢ରɻ͘ɺ҆͘ɺΑΓਖ਼֬ ৯ɿऩ֭ྔ૿Ճɺපݪੑɺ ্࣋ͪ ச࢈ཆ৩ɿےྔ૿ɺපݪੑ ҩྍɿༀɺҨࢠ࣬ױɺ ײછɺѱੑजᙾ
ྔతมԽ࣭తมԽΛੜΉ ྫɿ௨৴ճઢ ిใɿ1870’s ςΩετϝʔϧ Իి ΠϯλʔωοτʢςΩετɾը૾ɾԻָʣ ಈը৴ɿ1990’s ిใ ɿ ಈը৴
ʹ ࠓͷόΠΦҩༀ ɿ ʁʁʁ () (ߴ)
ɾݚڀͷͨΊͷݚڀ ͜Ε·Ͱ దͳ࣬ױϞσϧΛ ࡞Δͷʹ ʮ͜ͷϚε͕Ͱ͖Ε XXༀͷݚڀ͕Մೳʹ…ʂʯ CRISPR ελʔτϥΠϯ͕ ͔͜͜Βʹ Ϛε࡞࡞ۀ1
ݚڀޮ͕ܶతվળ ৽ༀ ΞϧπϋΠϚʔɺύʔΩϯιϯɺ ౷߹ࣦௐɺetc. ͷ࣏ྍༀʹظ..
ҩྍʹ͓͚ΔCRISPRͷՄೳੑ Ҩࢠ࣬ױ ײછ HIVʹײછ͠ͳ͍Tࡉ๔ʢന݂ٿʣΛ࡞ͬͯମɻ ߅HIVༀ͕ෆཁʹͳΔ࣏ྍޮՌ͋Γʢྟচࢼݧதʣ ҨࢠมҟΛम෮ͨ͠ࡉ๔Λମɻ ෆ͢Δ߬ૉϗϧϞϯΛੜ࢈ͯ͠පঢ়վળ ѱੑजᙾ ͕Μࡉ๔ʹແྗԽ͞Εͣɺ߅͕Μࡎ߅ੑ࣋ͭ Tࡉ๔Λ࡞ɻന݂පʹஶޮʢྟচࢼݧதʣ
BCI (Brain Computer Interface)
Brain Computer Interfaceͱ ͦͷͷΛσόΠεͱͯ͠͏ٕज़ ະདྷͷSFٕज़Ͱͳ͘ɺݱࡏਐߦܗͷςΫϊϩδʔ ͔ΒͷΞτϓοτ - 1998 αϧ͕ࢥߟͷΈͰΞʔϜΛૢ࡞ʢ1,000kmԕִʣ -
1999 ҙࢥද͕ࣔͰ͖ͳ͍ױऀ ʹΑΔจࣈձ࣮ݱ ͷΠϯϓοτ - 1961 ਓࣖͷୈҰྫ - 1968 ਓບͷୈҰྫ
σόΠεͷछྨ ২ࠐΈܕ ணܕ - ಄֖ࠎ෦ʹిۃ - ʹೖ·ͨද໘ཹஔ - ऑ͍ݕग़Մ -
खज़ෆཁͰ؆ศ - ײ͘ൃల్্ - ຖճௐ͕ඞཁ
2017্ظͷಈ͖ - ΠʔϩϯϚεΫઃ (Tesla, SpaceX) - ২ࠐΈܕBCI - ʮ݈߁ਓͷBCIΛ10Ҏ ʹ࣮ݱʯ
(2017/3) Google, Microsoft, Intelࣾʹ BCI ෦Λ༗͢Δ ΠʔϩϯϚεΫ࠷ऴతʹਓؒͱAIͷ༥߹ΛݟࠐΉ - ࢥߟʹΑΔλΠϐϯά σόΠε - ணܕBCI - ʮ100ޠ/minͷػثΛ 2Ͱ࡞ʯ (2017/4)
ʮࢥߟ͕Εͯ͠·͏ͷͰʁʯ ͷ͋ΒΏΔిؾత׆ಈ ࢥߟɾҙਤ ཧ ײ ໋ྩɾࢦࣔ ʮΛಈ͔ͤʯ ʮΛग़ͤʯ ʮจࣈΛଧͯʯ ग़ྗର͜ͷϨΠϠʔɻ
ͷલͷਓΛ - ମ͕উखʹԥͬͯ͠·Θͳ͍ - ແҙࣝʹഃͯ͠͠·Θͳ͍ ͷͱಉ͡Α͏ʹ੍ࣗՄೳ ͜ͷϨΠϠʔ͕͖উखʹ ग़ྗ͞ΕΔΘ͚Ͱͳ͍
ϩϘοτΞʔϜͷΈ ਓؒޠͷΛػցֶश͠ɺػցޠʹࣗಈ༁ͯ͠ૢ࡞ खΑಈ͚ɺखΑಈ͚ɺख Αಈ͚ɺखΑಈ͚ɺखΑ ಈ͚ɺखΑಈ͚ɺखΑ… खΛಈ͔࣌͢ͷ ࢲͷҰཡ 1. ʙʙ 2.
ʙʙ … ׳ΕΕɺ೦͡Δ ඞཁ͢Βແ͘ͳΔ
- ࢹਆܦେΛܹ͢Δ͜ͱՄೳ ϚΠΫ͔ΒௌਆܦɺΧϝϥ͔ΒບΛܹ ਓࣖɺਓບ - ੜདྷશͷਓ͕ਓࢹ֮ͰॳΊͯݟΔੈքଞͱಉ͔͡ʁ - ͷਓ͕ڹఆҐͰʮݟΔʯੈքͲ͏ݟ͑Δͷ͔ʁ - ࢲͷײ͡Δʮʯͱ͋ͳͨͷʮʯಉ͡৭͔ʁ
ʮݟ͑ํɾௌ͑͜ํʯҰ௨ΓͰͳ͍ ੈքͱΠϯλϥΫτͰ͖ΕɺଞਓͱҧͬͯͦΕ֮
ݟΔɺௌ͘ɺಈ͔͢ – ͷಇ͖ ʹύλʔϯΛ֮͑ࠐ·ͤΔ͜ͱ͕ຊ࣭ ຊޠʹஞޠ༁ͯ͠ཧղ ʢӳձͷशख़ʣ ѫࡰఔύλʔϯೝࣝͰ ༁ͣ͞ʹ͑Δ ෳࡶͳձ ӳޠͷ··ཧղͰ͖Δ
ෳࡶͳձ ӳޠͰߟ͑ͯൃͰ͖Δ ৗʹࢥߟΛհ͢Δ ʮ͜ͷܗʰυʱͩͬͨͳ…ʯ ʢ BCI ͷशख़ʣ ʮυϛι͚ͩϚελʔͨ͠ͳʯ ʢਓࣖͰͷύλʔϯೝࣝʣ ෳࡶͳܗࢥߟΛհͣ͞ ܗͷ··ཧղʢΠϯϓοτʣͰ͖Δ ෳࡶͳܗແҙࣝతʹ ग़ྗʢΞτϓοτʣͰ͖Δ
ҩྍʹ͓͚ΔBCIͷՄೳੑ ࣦΘΕͨΞτϓοτػߏͷ֫ಘ ࣦΘΕͨΠϯϓοτػߏͷ֫ಘ ࢹྗɺௌྗɺຯ֮ɺᄿ֮ɺ৮֮ ࢛ࢶͷಈ͖ɺίϛϡχέʔγϣϯ কདྷతʹ ӡಈೳྗɺײ֮ೳྗͷ֦ு ࡾຊͷखʢӡಈʣɺԻϏδϣϯʢײ֮ʣɺ खज़ϩϘοτࢹڸͱͷҰମԽʢӡಈ+ײ֮ʣ
None
ະདྷطʹ͜͜ʹ͋Δ ۉʹߦ͖Θ͍ͨͬͯͳ͍͚ͩͩ –– ΟϦΞϜɾΪϒιϯʢ࡞Ոʣ
ϒϩά http://okiyamasho.com ɾུྺ 2010 – ౦ژେֶҩֶ෦ଔۀ 2012 – ٹ໋ٹٸҩʢҩྍηϯλʔʣ 2013
– υΫλʔϔϦɺౡҩྍʢੴ֞ౡɺরؒౡʣ 2015 – גࣜձࣾϝυϨʔʢࣥߦһʣ 2017 – ϑϦʔʢݚڀऀ/ใֶʣ ɾॴଐֶձ ਓೳֶձɺใॲཧֶձɺຊٹٸҩֶձʢઐҩʣɺຊՊֶձ ԭࢁ ᠳʢ͓͖· ͠ΐ͏ʣ