$30 off During Our Annual Pro Sale. View Details »
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Without Practice, No Emergence
Search
Koichi ITO
June 26, 2023
Education
4
2.7k
Without Practice, No Emergence
Koichi ITO
June 26, 2023
Tweet
Share
More Decks by Koichi ITO
See All by Koichi ITO
STYLE
koic
0
220
Ruby and LLM Ecosystem
koic
3
6.2k
Write Code Every Day
koic
1
2.2k
Bliki (ja), and the Cathedral, and the Bazaar
koic
7
2.8k
Welcome to the LLM Club
koic
0
350
RuboCop: Modularity and AST Insights
koic
3
5.1k
Carving the Way to Ruby Engineering
koic
3
1.1k
Beyond the RuboCop Defaults
koic
3
4.9k
Minify Ruby Code
koic
2
2.5k
Other Decks in Education
See All in Education
ThingLink
matleenalaakso
28
4.2k
the difficulty into words
ukky86
0
330
1111
cbtlibrary
0
240
Evaluation Methods - Lecture 6 - Human-Computer Interaction (1023841ANR)
signer
PRO
0
1.2k
バケットポリシーの記述を誤りマネコンからS3バケットを操作できなくなりそうになった話
amarelo_n24
1
150
The World That Saved Me: A Story of Community and Gratitude
_hashimo2
3
430
HTML5 and the Open Web Platform - Lecture 3 - Web Technologies (1019888BNR)
signer
PRO
2
3.1k
2025年の本当に大事なAI動向まとめ
frievea
0
100
吉岡研究室紹介(2025年度)
kentaroy47
0
790
Презентация "Знаю Россию"
spilsart
0
380
Adobe Express
matleenalaakso
1
8.1k
今の私を形作る4つの要素と偶然の出会い(セレンディピティ)
mamohacy
2
130
Featured
See All Featured
Unsuck your backbone
ammeep
671
58k
How Fast Is Fast Enough? [PerfNow 2025]
tammyeverts
3
400
How to Create Impact in a Changing Tech Landscape [PerfNow 2023]
tammyeverts
55
3.2k
Marketing to machines
jonoalderson
1
4.3k
30 Presentation Tips
portentint
PRO
1
170
Designing for Performance
lara
610
69k
CoffeeScript is Beautiful & I Never Want to Write Plain JavaScript Again
sstephenson
162
16k
The Success of Rails: Ensuring Growth for the Next 100 Years
eileencodes
47
7.9k
Taking LLMs out of the black box: A practical guide to human-in-the-loop distillation
inesmontani
PRO
3
1.9k
How to train your dragon (web standard)
notwaldorf
97
6.4k
Exploring the relationship between traditional SERPs and Gen AI search
raygrieselhuber
PRO
2
3.4k
VelocityConf: Rendering Performance Case Studies
addyosmani
333
24k
Transcript
ࢲͷεϥΠυ࡞ज़ Without Practice, No Emergence !LPJD
w %&ʹڞ௨͍ͯ͠Δ֎෦ͰϓϨθϯ͢Δͱ ͍͏͜ͱʹ͍ͭͯ w ಘखෆಘखΑΓΔ͔Βͳ͍͔ΛΔ ʹ͖͔͚ͬͮ͘͢Γ w զʑͷࣄͱΩϟϦΞʹ͍ͭͯܨ͕Δ ࠓͷͳ͠
IUUQTFTNDPKQFTN@QFPQMFJOTJEF ࣗݾհ શࣾ)1ΑΓ
લ࢙ ੈلʙθϩલ IUUQTFTNDPKQFTN@QFPQMFJOTJEF
ߨࢣ࣌
w ͱʹ͔͘ͷྲྀΕॏཁͰετʔϦʔΛߟ͑Δ w εϥΠυͷಡΈ্͛Ͱ͋ΕɺεϥΠυΛ ͢ΕεϐʔΧʔͷਓෆཁͱ͍͏ҙ w ͋͑ͯεϥΠυʹશ෦Λॻ͔ͣ5JQTతͳͷ ޱ಄Ͱ͑ΔͱɺࢀՃऀͷ͓ಘײΛ࡞Ε Δ ࣌ͷҹʹֶͬͨͼ
ӬೖࣾޙͷϓϨθϯػձ w 4MJEF4IBSFͷެ։εϥΠυ࡞ w 4QFBLFS%FDLͷެ։εϥΠυ࡞ w ͔ΒͷӬೖࣾपͷ ؒຖ݄ϓϨθϯ͢Δͱ͍͏اըΛͨ͠ w ࣄͰͷඇެ։ͳͲؚΊͱʹ͔͘ࢁ
w ࣄͰ͍ͯ͠Δ͜ͱϓϨθϯ͠ͳ͍ͱɺ ͋ͳͨͷࣄ୭ԿΘ͔Βͳ͍ w Φʔϓϯͳ044׆ಈʹ͓͍ͯ3VCZ,BJHJ ͰͷूେΛ·ͱΊͯ͢ w ΞδϟΠϧࣄۀ෦ͰΤϯδχΞͷٕज़ൃ ৴ͳͲͷϓϨθϯ͕Ӧۀ׆ಈ ΤϯδχΞͷࣗݾൃ৴࣌
w ίϛϡχςΟͱͷڞੜΛܝ͛Δ w 044׆ಈొஃࣗݾհΛ݉ͶΔ w ໊Βࣺͬͯͯ·͚͢ͲͶ w ։ൃऀͷإ͕ݟ͑Δ৫
ઐӦۀͷ͍ͳ͍෦ॺ
ࠓͷҰ
4PGUCBOL $SFBUJWF
Ͱ͔͍ϓϨθϯ ࢲ͋ͱ͕͖ʹνϣΠग़͖͠·ͨ͠
ߴڮϝιου IUUQXXXSVCZDPMPSPSHUBLBIBTIJ
"NB[POͰങ͑·͢😅
w εϥΠυຕ͕ଟ͍ w ޙΖͷਓͰಡΊΔϑΥϯταΠζ w ߴڮϝιου෦తʹ͏ ߴڮϝιουʹӨڹ͞Εͨࢲͷ࡞෩
εϥΠυ৬ਓͱͯ͠ͷӨڹ
Өڹ
w 1PXFS1PJOUʹͦΜͳ ៉ྷͳ ϑΥϯτ ͋ΔΜͩͶ w ,FZOPUFʹͦΜͳ ඒ͘͠ͳ͍ ϑΥϯτ ͋ΔΜͩͶ
w ϑΥϯτΛങ͏ͱ͍͏ൃ ϑΥϯτॏཁ
w େࡕʹ͋ΔߤۭձࣾͷϗςϧϥϯδͰߦ ΘΕͨॻ੶ͷήϥϋοΫͰLENTOS͞Μ͕ ײ৺͍ͯͨ͠ϑϨʔζͱهԱ ଓࢺ ͷΈͷεϥΠυ Θͳ͍
w લͷϖʔδʹͬͨεϥΠυΛ࠶දࣔ͠ ͍ͨΑ͏ͳͱ͖ɺ༧Ίίϐʔͯࠩ͠͠ࠐ ΜͰ͓͘ w ྫ͑ຕ·ͰਐΜͰ͔Βɺຕͷε ϥΠυΛදࣔ͢ΔετʔϦʔͷ߹ຕ ͷεϥΠυΛίϐʔͯ͠ຕʹೖΕΔ લͷεϥΠυʹ͞ͳ͍
w ͔ͨ͘ʹ͞Μʹڭ͑ͯΒͬͨςΫχοΫ w ͜ͷςΫχοΫ͚ͩͰεϥΠυͷ͕֨ άοͱ্͕Δ ը૾ಁաͱςΩετγϟυ
IUUQTLBLVUBOJDPNIUNMQ εϥΠυͮ͘Γͷύλʔϯީิ %SPQ:PVS-BTU4MJEFT
w ೖ೦ʹϦϋʔαϧ͢Δਓ࣮ࡍ͏·͍ w ͨͩࣗͦͷ͋ͨΓʹۈษͳਓͰͳ͍ ͷͰɺεϥΠυ࡞࣌ʹͷྲྀΕΛ܁Γฦ ͠ਪᏏͯ͠಄ʹೖΕͯऴΘΓ w ͭ·ΓϦϋ͠ͳ͍ Ϧϋʔαϧͷ༗ແ
ଟ͘ͷ࣮ྫΛΔ
w ςοΫΠϕϯτོͷݱ w 4QFBLFS%FDL4MJEF4IBSFͳͲʹΞο ϓϩʔυ͞ΕͨεϥΠυΛɺετʔϦʔ ߏσβΠϯͳͲݚڀରͱͯ͠ݟΔ w ςοΫΠϕϯτʹࢀՃͯ͠ࢁͷϓϨθϯ ΛݟΔ ΦϑϥΠϯొஃΛݟΔͷ͕͓નΊ
͍ͭͰͰ͖ΔϓϨθϯͷֶͼํ
ൃදωλ͕ͳ͍ʁ
w ࣌ؒಇ͍͍ͯͯɺԿωλ͕ͳ͍Θ͚ ͳ͍ w ΞΠσΞ͕߱Γͯ͘ΔλΠϛϯάͱൃදλΠ ϛϯάผͳͷͰɺͱʹ͔͘ࢥ͍͍ͭͨΞΠ σΞϝϞ͓ͯ͘͠ w ϝϞ͢Ε҆৺ͯ͠ΕΒΕΔCZUXBEB ωλΛετοΫ͢Δ
w ߱Γ͖ͯͨΞΠσΞͷͷߏʹ͢ɺε ϥΠυߏΛࢥ͍͍ͭͨΒϝϞΛΞοϓσʔτ ͢Δ w πʔϧ(NBJMͷԼॻ͖Ͱɺ#BTFDBNQͦ ͷଞͳΜͰྑ͍ w ϝϞͨ͠ઌΛΕͳ͍ීஈ͍ͷπʔϧ͕ྑ͍ ωλݩΛΞοϓσʔτ͢Δ
w ࢥ͍͍ͭͨߏϚʔΫμϯͳͲͰϝϞΛॻ͍ͯ ͓͘ͱɺ࣮ϓϨθϯςʔγϣϯπʔϧʹམͱ͠ࠐΉ ͱ͖ʹศར w 3BCCJULLVCVONELFZΛ͏ͱͦͷ··ຊ ൪ϓϨθϯͰ͖Δ w IUUQTHJUIVCDPNSBCCJUTIPDLFSSBCCJU w
IUUQTHJUIVCDPNLLVCVONELFZ ߏΛϝϞ͓ͯ͘͠ͱศར
ϓϩϙʔβϧهड़ ͷίπʁ
৹ࠪһ͕ϫΫϫΫ ͢ΔΑ͏ͳ༰Λ ॻ͘ͱྑ͍
εϥΠυΛ࡞Δ ͕࣌ؒͳ͍ʁ
͑Δ͕͖࣌ؒ͞ɺ ظͷௐ͋ͱ
ҰຕͰҰจ͚ͩ IUUQTLPJDIBUFOBCMPHDPNFOUSZPNPUFTBOEPSCUI w ొஃͳΒۭ͍͍ͯͨ w εϥΠυຕͰޠΕΔͷ w ಥʹ.BU[ࢀՃͷձͩͬͨ
εϥΠυͳ͠ қ͕ߴ͍
ൃද࣌ͷίπʁ
w Կਓ͔͕ൃද͢ΔܗࣜͷϓϨθϯͷ߹ʹ ɺલޙͷൃදऀɺ͋Δ͍͕ࣗҹʹ ͬͨϓϨθϯʹ͍ͭͯݴٴ͢ΔͱϥΠϒ ײ͕૿͢ w લ·Ͱ࡞͍ͬͯΔͱ͜Ε͕ͮ͠Β͍ͷ ͰɺΏͱΓ͓࣋ͬͯ͘ͷ͕͓͢͢Ί લޙͷਓͷൃදʹ৮ΕΔ
w Ή͠Ζʮ͜͜ޙͰ ͓ͯ͜͠͏ʯͱݴͬ ͓͚ͯྑ͍ w ϥΠϰͰमਖ਼͢Δͷ ͻͱͭͷख w ͪΖΜेηϧϑ ϨϏϡʔ͢Δ
UZQP͕͋ͬͯؾʹ͠ͳ͍ IUUQPCKFDUDMVCKQEPXOMPBEpMFTJQQPO"JSQEG ΠϯλϥΫςΟϒϓϨθϯςʔγϣϯύλʔϯζ
͓ΘΓʹ
ׂͷ༏ΕͨਓʹίετΛ͔͚ͨ΄͏͕ɺׂ ͷී௨ͷਓΛఈ্͛͢ΔΑΓ͕૿͑ΔͷͰ ͳ͍͔ ʮී௨ͷਓʯׂ͕͍Δ͔Βʹɺී௨ͷϗ τ͕ී௨Ҏ্ʹͳΕΔ΄͏ʹίετΛ͔͚ͨ΄ ͏͕ɺશମͱͯ͠ͷͷ૯ྔେ͖͘ͳΓ·͢ ׂׂ͔͔ ݁ͷͳ͍ ʰ͍͢͝ϓϨθϯߴڮϝιουͷຊʱϖʔδΑΓ
8JUIPVU1SBDUJDF /P&NFSHFODF मͤ͟ΕݱΘΕͣ