AI and Design Vol. 2: Navigating the New Frontier - Morgenbooster
0 likes126 views
The document discusses the integration of AI in creativity and design, emphasizing its potential benefits while cautioning against over-reliance on automation. Key insights from various industry leaders highlight the importance of understanding AI's capabilities and using it to augment human creativity rather than replace it. The content also covers the evolution of open-source AI projects and their impact on the creative landscape, while citing the need for ethical considerations and user experience enhancement.
AI and Design Vol. 2: Navigating the New Frontier - Morgenbooster
- 2. AI a nd Desi g n. N a vi g a t in g t he New f ron t ier. Mor g enboos t er Copenh a g en, M a rch 22nd 2024
- 3. A l i t t l e In t roduc t ion Thom a s A l ex a nder Woo l f f Cre a t ive Techno l o g is t D a nie l Win t her-Korn Senior Desi g ner
- 4. Wh a t we’ l l be t a l kin g a bou t Thom a s A l ex a nder Woo l f f Scien t i f ic pro g ress wi t h AI Open Source AI Pro j ec t showc a se C l osin g rem a rks D a nie l Win t her-Korn A l i t t l e b a ck g round Wh a t a re we doin g wi t h AI The po t en t i a l o f AI & Cre a t ivi t y A f ew ( f u t ure) ex a mp l es
- 5. Am I bein g t hre a t ened b y a Robo t ? In a previous l y unrepor t ed quo t e, S a m A l t m a n, CEO o f OpenAI s a y s t h a t A l wi l l do “…95% o f wh a t m a rke t ers use a g encies, s t r a t e g is t s, a nd cre a t ive pro f ession a l s f or t od a y ."
- 12. – Rick Rubin “If you know what you want to do and you do it, that’s the work of a cra ft sman. If you begin with a question and use it to guide an adventure of discovery, that’s the work of the artist. The surprises along the way can expand your work, and even the art form itself.” The Cre a t ive Ac t : A W a y o f Bein g
- 14. AI c a n be t he so l u t ion AI c a n in f orm t he so l u t ion
- 15. AI c a n be t he so l u t ion
- 16. AI c a n in f orm t he so l u t ion work.co/clients/planned-parenthood/
- 19. Exp l ore more. And F a s t er! “…To be evaluated, ideas have to be seen, heard, tasted, or touched.” – Rick Rubin
- 21. A c a se: “never work wi t h chi l dren or a nim a l s”
- 29. Sudowri t e
- 31. - G e t experimen t in g - I’d recommend im a g e g ener a t ion, i t ’s f un! - The j ourne y is a s impor t a n t a s t he des t in a t ion - Fo l l ow some peop l e… A f ew t hou g h t s x.com/MartinNebelong Martin Nebeling x.com/juliewdesign_ Julie Wieland x.com/nickfloats Nick St. Pierre
- 32. Scien t i f ic pro g ress wi t h A I Open Source A I Pro j ec t showc a se C l osin g rem a rks
- 33. Scien t i f ic pro g ress wi t h AI Machines for the betterment of peop l e and p l anet rather than just to y s.
- 35. A l pha G o I t’s a l l fun and games unti l the robot beats y our ass! deepmind.google/technologies/alphago
- 38. G NoME G raph Networks for Materia l s Exp l oration deepmind.google/discover/blog/millions-of-new-materials-discovered-with-deep-learning 2.2 mi l l ion 800 y ears 52.000 l a y ered compounds 528 l ithium ion conductors
- 39. G NoME G raph Networks for Materia l s Exp l oration nature.com/articles/s41586-023-06735-9 736 41 in 17 hours
- 40. Open Source A I Benefits and l imits. The Wh y and the how?
- 42. " I f m y ca l cu l ations are correct, when this bab y hits 88 mi l es per hour, y ou're gonna see some serious shit.” Time l ine Dr. Emmett Brown
- 43. Time l ine Meta re l eased LLaMA to the research communit y under a noncommercia l l icense. Within a week of LLaMA's re l ease, it were l eaked to the pub l ic and thus ignited the open source LLM deve l opment February 2023 Mistra l re l eases their 7B mode l September 2023 Mistra l re l eases their Mixtra l of experts mode l that outperforms Chat G PT 3.5 December 2023
- 44. Time l ine Mistra l Large is re l eased which is a mere 5.2% behind Chat G PT 4. A l though this is not a open source mode l , but a paid version, it bui l ds upon the ear l ier Mistra l open source mode l s. February 2024 Over 555,000 Open Source mode l s shared on hugginface.co. Now a l l the major A I Deve l opers have some kind of open source product. March 2024
- 45. Bene f i t s Limi t s Accessibi l it y and Cost Efficienc y Customisation and F l exibi l it y Communit y Support and I nnovation Transparenc y and Trust I nteroperabi l it y and I ntegration Se l f hosting Sustainabi l it y Resource I ntensit y Maintenance and Support Cha l l enges Qua l it y and Re l iabi l it y Variations
- 47. • I ns t a l l pre- t r a ined mode l • Cre a t e d a t a se t (shi t in, shi t ou t ) • Lo a d mode l • Lo a d t okenizer • Add a dop t er • Tr a in mode l • S a ve mode l • Ev a l u a t e mode l • Run I n f erence • Push t o hos t (c a n be hu g g in f a ce.co or own server) • Use (s t a r t bui l din g y our f ron t end t o in t er a c t wi t h i t ) mlabonne.github.io/blog A speed y how- t o- g uide
- 50. A cautionar y ta l e of focus and buzz Web//:SizeIndex
- 51. • Nov. 2023 - ChatGPT Assistants announced. I created my Web Sustainability Assistant and ran first tests. The idea was to create a personal web sustainability audit based on the users website, by pairing Lighthouse data with Web Sustainability Guidelines. Could AI create a relevant and relatively easy to understand output with recommendations on how to make the users site more sustainable? • Dec. 2023 - Started building the tool as a side project using Next.js, PageSpeedInsights API and OpenAI API. And did ALOT of research on calculating co2 emissions by data transfer and consumption. • Jan. 2024 - Expanded with APIs for checking host energy type, co2 emissions, electricity price and added a database to store the data collected. • Feb. 2024 - OpenAI dropped from project. Fine-tuned Mistral dropped too. Decided to build the recommendations engine myself. • Mar. 2024 - Design, UX and Client Lead added to project. AI work starts. Can we successfully use AI to do a simple change to the UI? w3c.github.io/sustyweb Web//:SizeIndex
- 52. Learnings Start sma l l . Don’t use techno l og y for techno l og y ’s sake. Keep focus. Don’t l et the buzz derai l y ou.
- 54. C l osin g Rem a rks A parting gift of a bunch of recommendations
- 55. John M a ed a ’s 3 t a ke a w a y s f or SXSW 2024 (VP of Design and AI at Microso ft )
- 56. Compu t a t ion a l t hinkin g is inv a l u a b l e Coding isn’t ke y - but understanding computation wi l l he l p. Learn how to speak machine.
- 57. Work t r a ns f orm a t ion is comin g FAST Prep a re f or A I -induced shi f t s in desi g n a nd t ech c a reers, con t inu a l l y . I t ’s a m a r a t hon, no t a sprin t .
- 58. Be cri t ic a l o f A I ( f or cus t omers) Eva l uate A I ’s impact critica l l y ; consider fairness and inc l usivit y . Show customers the benefits of responsib l e A I .
- 60. A l pha G o To y A l phaFo l d Machine for the betterment of peop l e and p l anet Start sma l l and expand
- 61. Ask y ourse l f: Can we achieve the same (or c l ose) without it? G enerative A I
- 62. Adopt a Design Thinking approach to A I Human Centered: From the outside in. Ask y ourse l f: Does it enhance the users experience?
- 63. The 1508 Cre a t ive Tech 2024 A I s t r a t e g y Use A I to affect the U I rather than just creating another chat bot
- 64. TheTuringTrap direct.mit.edu/daed/article/151/2/272/110622/The-Turing-Trap-The-Promise-amp-Peril-of-Human The Promise & Peri l of Human-Like Artificia l I nte l l igence
- 65. “The future is not preordained. We contro l the extent to which A I either expands human opportunit y through augmentation or rep l aces humans through automation. We can work on cha l l enges that are eas y for machines and hard for humans, rather than hard for machines and eas y for humans.” direct.mit.edu/daed/article/151/2/272/110622/The-Turing-Trap-The-Promise-amp-Peril-of-Human Erik Brynjolfsson Professor and Senior Fellow at the Institute for Human-Centered AI and Director of the Digital Economy Lab at Stanford University
- 66. T a sks t h a t hum a ns c a n do Hum a ns t a sks t h a t m a chines c a n a u t om a t e New t a sks t h a t hum a ns c a n do wi t h t he he l p o f m a chines
- 67. “Techno l o g y ch a l l en g es us t o a sser t our hum a n v a l ues which me a ns t h a t f irs t o f a l l , we need t o f i g ure ou t wh a t t he y a re.” Sherry Turkle Professor of Science, Technology, and Society at MIT
- 68. Le t 's prove S a m A l t m a n wron g a nd not l e t A l a u t om a t e 95% o f our j obs. Bu t r a t her f ocus on usin g AI t o a u g men t t he work we do a nd co-cre a t e
- 70. Th a nk y ou
- 71. AI a nd Desi g n. N a vi g a t in g t he New f ron t ier. Q & A’s Mor g enboos t er Copenh a g en, M a rch 22nd 2024