Tweet
Tweet

Slide 1

Slide 1 text

Human-Informed Machine Learning Models and Interactions Hiromu Yakura Max-Planck Institute for Human Development IBIS 2024

Slide 2

Slide 2 text

 w ໼૔େເʢ΍͘ΒͻΖΉʣ w ϚοΫεϓϥϯΫਓؒ։ൃݚڀॴ ത࢜ݚڀһ w ͜Ε·Ͱʹ(PPHMF.JDSPTPGU3FTFBSDI1I%'FMMPX w σβΠϯࢧԉ͔Βൃୡো͕͍ࣇࢧԉ·Ͱ ػցֶशͷԠ༻Λ֦͛Δݚڀʹैࣄ w ઌ݄ΑΓ+45͖͕͚͞ʮࣾձมֵج൫ʯʹͯ ʮػցֶश࣌୅ͷࣾձม༰Λཧղ͢Δ ج൫Ξϓϩʔνͷ૑ग़ʯͱ͍͏՝୊Λ։࢝ ࣗݾ঺հ

Slide 3

Slide 3 text

ػցֶशͷ׆༻ͷͨΊͷ ΠϯλϥΫγϣϯݚڀ  ػցֶशͱΠϯλϥΫγϣϯͷؔ܎ੑ ػցֶशٕज़ͷݚڀ Ϣʔβ ϦεΫ΍ ݶք ࣾձͰͷ ड༰ w ػցֶशͷ࣮Ԡ༻ʹ޲͚ͯɺͦͷΪϟοϓΛຒΊΔ w ଓʑͱ࢈·ΕΔ৽ͨͳٕज़ΛͲ͏͢Ε͹࢖͍΍͘͢ͳΔ͔ ࣮ࡍͷιϑτ΢ΣΞ։ൃΛ௨ͯ͠ݕূ͠ɺํ๏࿦Λಋ͘ w ಛʹʮਓؒʯΛग़ൃ఺ʹλεΫΛଊ͑Δ͜ͱ͕ଟ͍

Slide 4

Slide 4 text

ػցֶशͷ׆༻ͷͨΊͷ ΠϯλϥΫγϣϯݚڀ  ػցֶशͱΠϯλϥΫγϣϯͷؔ܎ੑ ػցֶशٕज़ͷݚڀ Ϣʔβ ϦεΫ΍ ݶք ࣾձͰͷ ड༰ w ػցֶशͷ࣮Ԡ༻ʹ޲͚ͯɺͦͷΪϟοϓΛຒΊΔ w ଓʑͱ࢈·ΕΔ৽ͨͳٕज़ΛͲ͏͢Ε͹࢖͍΍͘͢ͳΔ͔ ࣮ࡍͷιϑτ΢ΣΞ։ൃΛ௨ͯ͠ݕূ͠ɺํ๏࿦Λಋ͘ w ಛʹʮਓؒʯΛग़ൃ఺ʹλεΫΛଊ͑Δ͜ͱ͕ଟ͍ """* *+$"*

Slide 5

Slide 5 text

Self-Supervised Contrastive Learning for Singing Voices Hiromu Yakura†‡, Kento Watanabe‡, Masataka Goto‡ † University of Tsukuba, Japan ‡ AIST, Japan IEEE/ACM Transactions on Audio, Speech, and Language Processing 2022 

Slide 6

Slide 6 text

 w ՎखࣝผɿՎ੠͔ΒରԠ͢ΔՎखΛࣝผ͢ΔλεΫ w Վ੠͔ΒͷՎखݕࡧͳͲʹԠ༻͢Δ͜ͱ͕Ͱ͖Δ w ػցֶशΛ༻͍ͨΞϓϩʔν͕෯޿͘༻͍ΒΕ͖͕ͯͨ ͦͷੑೳ͸ֶशʹ༻͍Δσʔληοτͷ࣭ʹେ͖͘ґଘ͢Δ w Վख৘ใ΍Վ੠ͷੑ࣭͕ద੾ʹΞϊςʔγϣϯ͞Εͨ େن໛σʔληοτΛ४උ͢Δͷ͸ϋʔυϧ͸ߴ͍ ՎखࣝผλεΫʹ͓͚Δσʔληοτ্ͷ੍໿ ࣗݾڭࢣ͋ΓֶशʹΑͬͯಛ௃ྔදݱΛ֫ಘ͢Δ͜ͱͰ ͦ͏ͨ͠σʔληοτ΁ґଘ͠ͳ͍ख๏Λ࣮ݱ͍ͨ͠

Slide 7

Slide 7 text

 w ࣗݾڭࢣ͋Γରরֶश< > ͸ϥϕϧͳ͠ͷσʔληοτ͔Β ಛ௃ྔදݱΛ֫ಘͰ͖ɺը૾υϝΠϯͰ޿·Γͭͭ͋Δ ࣗݾڭࢣ͋ΓରরֶशʹΑΔಛ௃ྔදݱͷ֫ಘ [5] Jaiswal, A. et al.: A survey on contrastive self-supervised learning, Technologies (2021). [6] Jing, L. and Tian, Y.: Self-supervised visual feature learning with deep neural networks: A survey, IEEE Trans. Pattern Anal. Mach. Intell. (2021). w ػցతʹม׵ͨ͠ೖྗର͕ ࣅͨજࡏදݱʹͳΔΑ͏ʹ ਂ૚ֶशϞσϧΛ܇࿅͢Δ

Slide 8

Slide 8 text

 w ࣗݾڭࢣ͋Γରরֶश< > ͸ϥϕϧͳ͠ͷσʔληοτ͔Β ಛ௃ྔදݱΛ֫ಘͰ͖ɺը૾υϝΠϯͰ޿·Γͭͭ͋Δ ࣗݾڭࢣ͋ΓରরֶशʹΑΔಛ௃ྔදݱͷ֫ಘ [5] Jaiswal, A. et al.: A survey on contrastive self-supervised learning, Technologies (2021). [6] Jing, L. and Tian, Y.: Self-supervised visual feature learning with deep neural networks: A survey, IEEE Trans. Pattern Anal. Mach. Intell. (2021). ୯७ʹը૾ͱಉ͡ܗͰ ϞσϧΛ܇࿅ͯ͠΋ ਫ਼౓͸શ͘޲্ͤͣ w ػցతʹม׵ͨ͠ೖྗର͕ ࣅͨજࡏදݱʹͳΔΑ͏ʹ ਂ૚ֶशϞσϧΛ܇࿅͢Δ

Slide 9

Slide 9 text

 w Վ੠ͷੑ࣭͸੠࣭ͱՎ͍ํʹ෼ղͯ͠ߟ͑Δ͜ͱ͕Ͱ͖Δ w ੠࣭͸εϖΫτϧแབྷ΍ϑΥϧϚϯτʹґଘ͢Δ w Վ͍ํ͸Ϗϒϥʔτ΍ΞʔςΟΩϡϨʔγϣϯʹݱΕΔ Վ੠ͷߏ੒ཁૉͱͦͷ੠࣭ ָثతͳ੠࣭ ʷ Ͳͷఔ౓λϝΔ͔ ϏϒϥʔτͷՃݮ ͦΕͧΕͷදݱʹΑΔՎ͍ํ ੠ಓͷܗ ͳͲʹ༝དྷ ʹ ࠷ऴతͳՎ੠ ͜ΕΒ͕ʮࣅ͍ͯΔʯͱ͸Ͳ͏͍͏ঢ়ଶ͔Λߟ͑ͯ ࣗݾڭࢣ͋ΓରরֶशΛઃܭ͢Δ

Slide 10

Slide 10 text

 w ͦ΋ͦ΋ʮ੠࣭͕ࣅ͍ͯΔʯͱ͸Ͳ͏͍͏͜ͱ͔ ੠࣭͕ࣅ͍ͯΔͱ͸Ͳ͏͍͏͜ͱ͔ ฏҪݎ ಏΛดͯ͡ Ұ੨ᜫʢϐονˣʣ ϋφϛζΩʊ Ұ੨ᜫ ϋφϛζΩ ग़య: https://dic.nicovideo.jp/id/330383 ɹɹ https://ameblo.jp/tmj-blog/entry-12344714192.html

Slide 11

Slide 11 text

w ͦ΋ͦ΋ʮ੠࣭͕ࣅ͍ͯΔʯͱ͸Ͳ͏͍͏͜ͱ͔ w ػցతʹϐονΛม͑Δͱશવҧ͏੠࣭ʹฉ͑͜Δ w ָثͷԻ৭ͱಉ༷ʹɺ੠࣭͸ प೾਺ଳͷ෼෍Ͱଊ͑ΒΕΔ w ಉ͡ਓ͕ҧ͏ߴ͞ͷԻΛ Վͬͯ΋େ·͔ͳࢁͷܗ͸ෆม w ػցతʹϐονΛม͑Δͱ ࢁͷܗ΋มΘͬͯ͠·͏  ੠࣭͕ࣅ͍ͯΔͱ͸Ͳ͏͍͏͜ͱ͔ ϐʔΫ ʢॎ๮ʣ ͷ ִؒ͸Ұఆ ػցతͳϐον ม׵ͷΠϝʔδ ग़య: S. Duvvuru, et al.: The Effect of Timbre, Pitch, and Vibrato on Vocal Pitch-Matching Accuracy. Journal of Voice (2016).

Slide 12

Slide 12 text

 Վ͍ํ͕ࣅ͍ͯΔͱ͸Ͳ͏͍͏͜ͱ͔ [31] Kako, T. et al.: Automatic identification for singing style based on sung melodic contour characterized in phase plane, ISMIR (2009). ػցతͳλΠϜ ετϨονͷΠϝʔδ w Ͱ͸ʮՎ͍ํ͕ࣅ͍ͯΔʯͱ͸Ͳ͏͍͏͜ͱ͔ w ඍখ࣌ؒ಺ʹൃੜ͢ΔΞʔςΟΩϡϨʔγϣϯʹண໨ w Իߴͷ੾ΓସΘΓͰͷ ' ͷޯ഑͸ ݸਓͷՎ͍ํ͕൓ө͞ΕΔ<>  w ಉ͡ਓ͕Ώͬ͘ΓՎͬͨ৔߹΋ ' ͷಈ͖ํ͸มΘΒͳ͍ w ػցతʹλΠϜετϨον͢Δͱ ' ͷࡉ͔ͳޯ഑ͷܗ΋มΘͬͯ͠·͏

Slide 13

Slide 13 text

 ࣗݾڭࢣ͋ΓֶशʹΑΔՎ੠ελΠϧͷಛ௃ྔදݱͷ֫ಘ w ͜ΕΒΛ൓ө͠ɺՎ੠ʹಛԽͨ͠ಛ௃ྔදݱΛ֫ಘ͢Δ ࣗݾڭࢣ͋Γରরֶशͷख๏ΛఏҊ ੠࣭΍Վ͍ํͷҧ͍Λ หผ͢Δಛ௃ྔදݱʹ ͳΔͱߟ͑ΒΕΔ ػցతʹϐονγϑτ΍ λΠϜετϨονͨ͠ Վ੠͸ผϞϊͱͯ͠ѻ͏

Slide 14

Slide 14 text

 w ·ͣՎख৘ใͳ͠ͷՎ੠σʔλͰࣗݾڭࢣ͋ΓֶशΛ࣮ࢪ w   ۂ ʷ  ඵͷԻָ %# Λɺ Վ੠෼཭<> ͯ͠܇࿅ʹ࢖༻ w ΞʔςΟετ໊΍ͦͷଞϝλσʔλ͸܇࿅ʹؚΊͣ w ্هͱ͸ผʹɺ ໊ ʷ  ۂͷσʔληοτΛߏங͠ μ΢ϯετϦʔϜλεΫͱͯ͠Վखࣝผͷਫ਼౓Λൺֱ w طଘख๏<> ͔ΒQU ఔͷ ਫ਼౓޲্Λ֬ೝ w Վ੠ͷੑ࣭Λ׆͔ͨ͠ ઃܭͷ༗ޮੑ΋֬ೝ ՎखࣝผλεΫͰͷධՁ [35] JHennequin, R. et al.: Spleeter: A fast and efficient music source separation tool with pre-trained models, J. Open Source Softw. (2020). [12] Spijkervet, J. and Burgoyne, J. A.: Contrastive learning of musical representations, ISMIR (2021).

Slide 15

Slide 15 text

 ੠࣭ʹಛԽͨ͠ಛ௃ྔදݱ Վ͍ํͷҧ͍͸ແࢹ͠ ੠࣭ͷҧ͍Λଊ͑Δ ಛ௃ྔදݱ͕ಘΒΕΔ λΠϜετϨονͨ͠ Վ੠͸ಉ͡΋ͷɺ ϐονΛม͑ͨՎ੠͸ ผ΋ͷͱͯ͠ѻ͏ w ม׵ͷ࢖͍ํΛ૊Έସ͑ͯɺఏҊख๏Λ֦ு͢Δ͜ͱ΋Մೳ w ྫ͑͹ɺՎ͍ํʹؔ܎ͳ͘੠࣭͚ͩͷྨࣅ౓Λࢉग़Ͱ͖Δ

Slide 16

Slide 16 text

 Վ͍ํʹಛԽͨ͠ಛ௃ྔදݱ ੠࣭ͷҧ͍͸ແࢹ͠ Վ͍ํͷҧ͍Λଊ͑Δ ಛ௃ྔදݱ͕ಘΒΕΔ λΠϜετϨονͨ͠ Վ੠͸ผ΋ͷɺ ϐονΛม͑ͨՎ੠͸ ಉ͡΋ͷͱͯ͠ѻ͏ w ม׵ͷ࢖͍ํΛ૊Έସ͑ͯɺఏҊख๏Λ֦ு͢Δ͜ͱ΋Մೳ w ٯʹɺ੠࣭ʹؔ܎ͳ͘Վ͍ํ͚ͩͷྨࣅ౓΋ࢉग़Ͱ͖Δ

Slide 17

Slide 17 text

 w ਓؒͷՎ੠ͷಛੑʹண໨͢Δ͜ͱͰ ࣗݾڭࢣ͋ΓରরֶशʹΑΔ৽ͨͳಛ௃ྔ֫ಘख๏ΛఏҊ w ϐονγϑτͱλΠϜετϨονΛޮՌతʹ༻͍Δ͜ͱͰ ՎखࣝผλεΫͷਫ਼౓Λେ෯ʹ޲্ͤ͞ΒΕΔ͜ͱΛ֬ೝ w ੠࣭ɾՎ͍ํͷ͍ͣΕ͔ͷΈʹ͍ͭͯͷྨࣅ౓Λଊ͑ΔΑ͏ ఏҊख๏Λ֦ு͢Δ͜ͱ΋Ͱ͖Δͱ֬ೝ w ྫ͑͹ɺՎ͍ํ͸·ͩ·͕ͩͩ੠࣭͕ࣅ͍ͯΔ ࠽ೳͷݪੴΛݟ͚ͭΔͱ͍ͬͨํ޲ੑ΋ʜʁ ·ͱΊ Ϣʔβͷଆ͔Β໰୊Λଊ͑௚͢͜ͱͰ طଘͷֶशख๏Λ֦େͯ͠৽ͨͳ࢖͍ಓΛੜΈग़͢

Slide 18

Slide 18 text

Tool- and Domain-Agnostic Parameterization of Style Transfer E ff ects Leveraging Pretrained Perceptual Metrics Hiromu Yakura†, Yuki Koyama‡, Masataka Goto‡ † University of Tsukuba, Japan ‡ AIST, Japan IJCAI 2021 

Slide 19

Slide 19 text

 w ਂ૚ֶश͕͜Ε·Ͱʹͳ͔ͬͨΑ͏ͳίϯςϯπੜ੒Λ࣮ݱ w ಛʹελΠϧసҠ͸෯޿͍υϝΠϯͰख๏ͷఏҊ͕͋Δ എܠਂ૚ֶशʹΑΔελΠϧసҠख๏ͷ޿͕Γ ෩ܠࣸਅ <-J > ΦϦδφϧ ϦϑΝϨϯε ελΠϧΛసҠ ϝΠΫࣸਅ <$IBOH > ("/ Y. Li, et al. A Closed-form Solution to Photorealistic Image Stylization. ECCV 2018. H. Chang. et al. PairedCycleGAN: Asymmetric Style Transfer for Applying and Removing Makeup. CVPR 2018.

Slide 20

Slide 20 text

 w Ұํɺ૑࡞දݱͷπʔϧͱͯ͠͸޿͘࢖ΘΕΔʹ·ͩࢸΒͣ w ྫʣ1SJTNBͱ͍͏ࣸਅՃ޻ΞϓϦ͕େྲྀߦͨ͠΋ͷͷ ɹɹ͍·Ͱ͸*OTUBHSBNʹ໭͍ͬͯΔϢʔβ͕େ൒ എܠελΠϧసҠͷ૑࡞πʔϧͱͯ͠ͷ޿·Βͳ͍ར༻ ग़య: https://jp.techcrunch.com/2016/07/20/20160719prismagram/

Slide 21

Slide 21 text

 w ͳͥελΠϧసҠ͕૑࡞πʔϧͱͯ͠޿·Βͳ͍ͷ͔ʁ   w ࠷ॳ͔Β׬ᘳͳΰʔϧΠϝʔδΛ͍࣋ͬͯΔͷ͸كͰɺ ΅΍ͬͱͨ͠Πϝʔδ͔ΒσβΠϯΛ࢝ΊΔͷ͕΄ͱΜͲ w ༷ʑͳࢼߦࡨޡΛ܁Γฦ͢தͰɺηϨϯσΟϐςΟతʹ ϏϏοͱ͖ͨ΋ͷ͕ɺ݁Ռͱͯ͠׬੒඼ͱͳΔ w ݱࡏͷελΠϧసҠ͸ɺϫϯγϣοτͰਫ਼៛ͳ݁ՌΛੜΉ͕ ਓ͕ؒࡉ͔͘ࢼߦࡨޡ͠ͳ͕Β୳ࡧ͍ͯ͘͠༨஍͕ͳ͍ എܠελΠϧసҠͱզʑͷσβΠϯϓϩηεͱͷෆҰக ਓؒͷσβΠϯϓϩηε͸ԟʑʹͯ͠୳ࡧతͰ<5BMUPO > ͭͷ݁Ռ͚ͩΛग़ྗ͢Δख๏ͱೃછ·ͳ͍͔Β J. Talton, et al. Exploratory modeling with collaborative design spaces. ACM Trans. Graph. 28(5). 2009.

Slide 22

Slide 22 text

 എܠ୳ࡧతϓϩηεͱ &OEUP&OE ελΠϧసҠͷࠩ େ·͔ͳ ׬੒Πϝʔδ ୳ࡧతϓϩηε ҉͞ ίϯτϥετ Ϗωοτ ΦϦδφϧ ͜Ε͕ ͍͍͔΋ʂ &OEUP&OEͷελΠϧసҠ ҉͞ ίϯτϥετ Ϗωοτ ΦϦδφϧ ελΠϧͷ ϦϑΝϨϯε ࢼߦࡨޡ͠ͳ͕Β௚ײతʹ σβΠϯۭؒΛཧղ͍ͯ͘͠ ϫϯγϣοτͰ ϕετͳ݁Ռʹ ग़ձ͏ͷ͸ ೉͍͠ ʷ ʷ ʷ ͍͔ʹࣅͤΔ͔Λ໨ࢦ͢ख๏Ͱ ࢼߦࡨޡʹ͸޲͔ͳ͍

Slide 23

Slide 23 text

 w ࣅͤͨ݁ՌͰ͸ͳ͘ɺࣅͤΔํ๏Λڭ͑Δͱ͍͏ΞΠσΞ w DG࿝ࢠʮڕΛ༩͑ΔΑΓ௼ΓํΛڭ͑Αʯ ఏҊύϥϝτϦοΫͳม׵ʹΑΔελΠϧసҠͷ໛฿ ׳Ε਌͠Μͩπʔϧ಺ͰͷࣅͤํΛڭ͑ͯ͋͛Ε͹ Ϣʔβ͸͔ͦ͜Βࣗ༝ʹฤूɾ୳ࡧͰ͖Δ ͲͷϑΟϧλͱ ͲΜͳύϥϝλΛ ࢖͑͹ࣅͤΒΕΔ͔ ΦϦδφϧ ϦϑΝϨϯε సҠ݁Ռ ݱঢ় ϒϥοΫ ϘοΫε ݁ՌɺϢʔβ͕ Α͍ͱࢥ͑ͨ෺ ʷ ΦϦδφϧ ϦϑΝϨϯε సҠ݁Ռ ݁ՌɺϢʔβ͕ Α͍ͱࢥ͑ͨ෺ ఏҊ ೖྗ ೖྗ

Slide 24

Slide 24 text

 w ࣅͤͨ݁ՌͰ͸ͳ͘ɺࣅͤΔํ๏Λڭ͑Δͱ͍͏ΞΠσΞ w DG࿝ࢠʮڕΛ༩͑ΔΑΓ௼ΓํΛڭ͑Αʯ ఏҊύϥϝτϦοΫͳม׵ʹΑΔελΠϧసҠͷ໛฿ ͲͷϑΟϧλΛ࢖͍ɺ ͲΜͳύϥϝλʹ͢Ε͹ ελΠϧ͕ࣅΔ͔͕෼͔Δ ׳Ε਌͠Μͩπʔϧ಺Ͱͷࣅͤํ͕Θ͔Δͱ ͦΕΛ΋ͱʹࣗ༝ʹ୳ࡧ͢Δͷ΋༰қ ϦϑΝϨϯε ม׵݁Ռ ΦϦδφϧ ϦϑΝϨϯε ม׵݁Ռ ΦϦδφϧ

Slide 25

Slide 25 text

 w Ͱ͸Ͳ͏΍ͬͯࣅͤํʢύϥϝλ΍ม׵ʣΛٻΊΔ͔ʁ w ࠷దԽ໰୊ͱͯ͠ଊ͑ͯ΋ɺ ม׵݁ՌͱϦϑΝϨϯεͷྨࣅ౓Λ ௚઀ൺֱͰ͖ͣ໨తؔ਺͕࡞Εͳ͍  w ͭͷ伴ͱͳΔཁૉΛಋೖ w ("/ͷજࡏදݱʹΑΔ஌֮తई౓ w ϒϥοΫϘοΫε࠷దԽ ఏҊύϥϝτϦοΫͳม׵ʹΑΔελΠϧసҠͷ໛฿ ม׵݁Ռ ϦϑΝϨϯε ݩʑͷࣸਅ͕ҧ͏ͷͰ ͲΕ͘Β͍ࣅ͍ͯΔ͔ͷ ܭࢉػతͳ൑அ͕೉͍͠

Slide 26

Slide 26 text

 w ("/ʹΑΔελΠϧసҠͰ͸ɺ&ODPEFS͕ελΠϧͱ಺༰Λ ෼཭͢Δ͜ͱͰɺελΠϧͷΈͷసҠΛՄೳʹ͍ͯ͠Δ      w ֶशࡁΈϞσϧͷ&ODPEFSͰಘΒΕΔελΠϧͷજࡏදݱΛ ൺֱ͢Ε͹ɺͲΕ͘Β͍ࣅ͍ͯΔ͔ͷई౓ΛಘΒΕΔ w ͔͠΋ɺࣸਅ΍ϝΠΫελΠϧͳͲɺϞσϧ͑͋͞Ε͹ ෯޿͍ର৅Λ࠶ֶशͳ͠ʹѻ͏͜ͱ͕Ͱ͖Δ ఏҊ("/ͷજࡏදݱʹΑΔ஌֮తई౓ ࣅͨελΠϧͱΘ͔Δ

Slide 27

Slide 27 text

 w ϒϥοΫϘοΫε࠷దԽΛ༻͍ɺ஌֮తई౓͕ۙ͘ͳΔΑ͏ ϑΟϧλ΍ύϥϝλΛࣗಈతʹ୳ࡧ͍ͯ͘͠       w ϒϥοΫϘοΫε࠷దԽ͸༻͍Δม׵Λ੍໿͠ͳ͍ͨΊɺ *OTUBHSBN͚ͩͰͳ༷͘ʑͳπʔϧͰͷࣅͤํΛ୳ࡧՄೳ ఏҊϒϥοΫϘοΫε࠷దԽʹΑΔࣅͤํͷ୳ࡧ

Slide 28

Slide 28 text

 w *OTUBHSBN ΍ 4/08 ͳͲͷ "1* ͷͳ͍ΞϓϦͰ͋ͬͯ΋ "OESPJE&NVMBUPS ͱςετϥΠϒϥϦͰࣗಈతʹૢ࡞ w ֶशࡁΈ஌֮ϞσϧͱϕΠζ࠷దԽͰঃʑʹ͚͍ۙͮͯ͘ ࣮૷

Slide 29

Slide 29 text

 w ʮࣅͤͨ݁ՌͰ͸ͳ͘ɺࣅͤΔํ๏Λڭ͑Δʯͱ͍͏खஈͷ ࣮ݱՄೳੑΛ෩ܠࣸਅɾϝΠΫՃ޻ͷγφϦΦͰݕূ w Ϋϥ΢υϫʔΧʹΑΔୈࡾऀओ؍ධՁͰΫΦϦςΟΛධՁ ධՁ݁Ռ ϦϑΝϨϯε ΦϦδφϧ ఏҊख๏ ڠྗऀ" ڠྗऀ# ϦϑΝϨϯε ΦϦδφϧ ఏҊख๏ ڠྗऀ" ڠྗऀ#

Slide 30

Slide 30 text

 w ਓؒͷσβΠϯϓϩηεͷಛੑʹண໨্ͨ͠Ͱ ελΠϧసҠϞσϧΛطଘΞϓϦʹ૊Έ߹ΘͤΔख๏ΛఏҊ w ಛʹɺయܕతͳػցֶशλεΫͱਓؒͷߦಈͷ Ϊϟοϓʹண໨͢Δ͜ͱͰ৽ͨͳ໰୊ઃఆΛఏى w ʮࣅͤͨ݁ՌͰ͸ͳ͘ɺࣅͤΔํ๏Λڭ͑Δʯͱ͍͏ Ξϓϩʔνֶ͕शࡁϞσϧͷΈͰ࣮ݱͰ͖Δ͜ͱΛ֬ೝ ·ͱΊ Ϣʔβͷೝ஌ɾߦಈ͔Β໰୊Λଊ͑௚͢͜ͱͰ طଘͷֶशϞσϧͷ৽ͨͳϢʔεέʔεΛੜΈग़͢

Slide 31

Slide 31 text

Human Behavior-Informed ML ML-Informed Human Behavior 

Slide 32

Slide 32 text

Empirical evidence of Large Language Model's influence on human spoken communication Hiromu Yakura*, Ezequiel Lopez-Lopez*, Levin Brinkmann*, Ignacio Serna, Prateek Gupta, Iyad Rahwan *: equal contribution Max-Planck Institute for Human Development  arXiv 2409.01754

Slide 33

Slide 33 text

$IBU(15 ͷ EFMWF όΠΞε ‣ $IBU(15 ͸ͳ͔ͥ EFMWF ͱ͍͏୯ޠΛ࢖͍͕ͪͱ͍͏ݱ৅͕ ޿͘஌ΒΕͭͭ͋Δ ‣ ଞʹ΋ $IBU(15 ʹಛ௃తͳ ୯ޠ͕ൃݟ͞Εɺ࿦จ౳Ͱ ग़ݱස౓ͷ૿Ճ͕ࢦఠ͞ΕΔ  W. Liang, et al. Mapping the Increasing Use of LLMs in Scientific Papers. Proc. CoLM (2024). https://pshapira.net/2024/03/31/delving-into-delve/

Slide 34

Slide 34 text

ࣗ෼Ͱ΋ৼΓฦͬͯΈΔͱ ‣ ࣗ਎΋Α͘ $IBU(15 ͰӳޠϝʔϧͷจষͳͲΛఴ࡟ ‣ ͦͯ͠ɺ$IBU(15 Λ࢖͍ͬͯͳ͍ͱ͖ʹ  EFMWF ͱ͍͏୯ޠΛ࢖͏บ͕఻છ͍ͯ͠Δؾ͕ ‣ ੈքͷֶज़ػؔͷ:PV5VCFνϟϯωϧ͔Β෼Ҏ্ͷ ಈըສ݅Λऩू͠ɺॻ͖ى্ͨ͜͠͠ͰมԽΛ෼ੳ  Ծઆ$IBU(15ͷଘࡏ͕ਓؒͷ ݴޠίϛϡχέʔγϣϯΛม͑ͭͭ͋ΔͷͰ͸ͳ͍͔

Slide 35

Slide 35 text

݁Ռ 

Slide 36

Slide 36 text

EFMWF ͸Ի੠ίϛϡχέʔγϣϯͰ΋૿͍͑ͯͨ ‣ $IBU(15 ʹಛ௃తͳ ୯ޠ͸ (15 ͷެ։ޙʹ ߹੒ίϯτϩʔϧͱ ൺ΂༗ҙʹස౓͕૿Ճ ‣ ߹੒ίϯτϩʔϧΛ ༻͍͍ͯΔͨΊ ҼՌޮՌ΋ࣔࠦ 

Slide 37

Slide 37 text

ΞϑϦΧӳޠ͕ $IBU(15 ܦ༝Ͱશੈքʹ఻೻͍ͯ͠Δʁ ‣ ͔͠΋ɺ͜͏ͨ͠୯ޠ͸έχΞӳޠͳͲͰΑ͘࢖ΘΕ͍ͯΔ ‣ 0QFO"* ͕ΞϊςʔγϣϯΛ ҆Ձʹ֎஫ͨ݁͠Ռ͕શੈքʹ Өڹ͍ͯ͠Δͱ͍͏Մೳੑ΋  https://time.com/6247678/openai-chatgpt-kenya-workers/

Slide 38

Slide 38 text

ػցֶश࣌୅ͷ ਓؒཧղͷݚڀ @ ػցֶशͷ׆༻ͷͨΊͷ ΠϯλϥΫγϣϯݚڀ  ػցֶशͱΠϯλϥΫγϣϯͷؔ܎ੑ ػցֶशٕज़ͷݚڀ Ϣʔβ ϦεΫ΍ ݶք ࣾձͰͷ ड༰ w ػցֶशٕज़͕࣋ͭྗ͕ڧ͘ͳͬͨࠓͦ͜ ػցֶशͷϥετϫϯϚΠϧ໰୊ʢʹͲ͏ಧ͚Δ͔ʣ͕ॏཁʹ w ਓؒΛग़ൃ఺ʹλεΫΛଊ͑Δ͜ͱ͕ώϯτʹͳΔ͔΋ w ʮՊֶతڵຯʯʹجͮ͘ݚڀ΋໘ന͘ͳΓͦ͏ͳ༧ײ͕