Continuous Learning_Startup & Investment
2.44K subscribers
513 photos
5 videos
16 files
2.74K links
We journey together through the captivating realms of entrepreneurship, investment, life, and technology. This is my chronicle of exploration, where I capture and share the lessons that shape our world. Join us and let's never stop learning!
Download Telegram
๐Ÿ’ก AI ์ œํ’ˆ์˜ UX๋Š” ๋ฌด์—‡์ด ๋‹ฌ๋ผ์•ผํ• ๊นŒ? - (1) ์ž๋™์˜ ํ•จ์ •

Vrew, ๋คผํŠผ, ๋ฆด๋ฆฌ์ŠคAI ๊นŒ์ง€ 3๊ฐœ์˜ AI ์ œํ’ˆ์„ ๋งŒ๋“ค๋ฉฐ ๋ฐฐ์šฐ๊ณ  ๋А๊ผˆ๋˜ ๊ฒƒ๋“ค์„ ํ•˜๋‚˜์”ฉ ๊ณต์œ ํ•ด๋ณด๋ คํ•ฉ๋‹ˆ๋‹ค ๐Ÿ™‚

์ธ๊ณต์ง€๋Šฅ์˜ ๋†€๋ผ์šด ๋ฐœ์ „์œผ๋กœ, ์ •๋ง ๋งŽ์€ ์ธ๊ฐ„์˜ ์ง€์  ์ž‘์—…์„ โ€˜์ž๋™ํ™”โ€˜ํ•  ์ˆ˜ ์žˆ๋Š” ๊ธฐํšŒ๊ฐ€ ์—ด๋ฆฌ๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.
ํ•˜์ง€๋งŒ ์—ญ์„ค์ ์ด๊ฒŒ๋„, ํŒ€ ๋ฆด๋ฆฌ์Šค๋Š” ์ œํ’ˆ์˜ UX์— ๋Œ€ํ•ด ๋…ผ์˜ํ•  ๋•Œ โ€˜์ž๋™์˜ ํ•จ์ •โ€˜์— ๋น ์ง€์ง€ ๋ง์ž๋Š” ์ด์•ผ๊ธฐ๋ฅผ ๋งŽ์ด ํ•ฉ๋‹ˆ๋‹ค.

AI์—๊ฒŒ ์–ด๋–ค ์ผ์„ ์‹œ์ผฐ์„ ๋•Œ 100์ ์„ ํ•ด๋‚ด๋Š” ๊ฒฝ์šฐ๋Š” ๋งŽ์ง€ ์•Š์Šต๋‹ˆ๋‹ค. ๋Œ€๋ถ€๋ถ„ 70์  ์ •๋„์— ๊ทธ์น˜๊ณ  ๋ง™๋‹ˆ๋‹ค.
70์ ์œผ๋กœ ์ดˆ์•ˆ๋งŒ ๋งŒ๋“ค์–ด ์ค˜๋„ ์ถฉ๋ถ„ํžˆ ๋งŽ์€ ์‹œ๊ฐ„์„ ๋‹จ์ถ•ํ•ด์ฃผ๋‹ˆ ์˜๋ฏธ ์žˆ๋Š” ๊ฒƒ์ด ์•„๋‹๊นŒ? ๋ผ๊ณ  ์ƒ๊ฐํ•  ์ˆ˜๋„ ์žˆ๊ฒ ์Šต๋‹ˆ๋‹ค. ํ•˜์ง€๋งŒ ์‚ฌ์šฉ์ž์˜ ์›Œํฌํ”Œ๋กœ์šฐ๋ฅผ ๋”ฐ๋ผ๊ฐ€๋‹ค๋ณด๋ฉด, ๊ฒฐ๊ตญ ์›ํ•˜๋Š” ๋ชฉ์ ์„ ๋‹ฌ์„ฑํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” AI๊ฐ€ ๋งŒ๋“ค์–ด์ค€ 70์  ์งœ๋ฆฌ ๊ฒฐ๊ณผ๋ฌผ์„ ๋‹ค ๋ถ€์ˆ˜๊ณ  ์ฒ˜์Œ๋ถ€ํ„ฐ 100์„ ๋งŒ๋“ค์–ด์•ผ ํ•˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋นˆ๋ฒˆํžˆ ๋ฐœ์ƒํ•ฉ๋‹ˆ๋‹ค.
์ด๋Ÿฐ ์ œํ’ˆ์˜ ๊ฒฝ์šฐ ์‚ฌ์šฉ์ž๋“ค์ด ์ฒ˜์Œ์—” โ€˜์™€~ ์‹ ๊ธฐํ•˜๋‹ค~โ€˜๋ผ๊ณ  ๋ฐ˜์‘ํ•˜์ง€๋งŒ, ์‹ค์ œ๋กœ๋Š” ์“ธ๋งŒํ•œ ๋ ˆ๋ฒจ๊นŒ์ง€ ๊ฐ€์ง€ ๋ชปํ•ด ๋ฆฌํ…์…˜์ด ์ž˜ ์•ˆ ๋‚˜์˜ค๊ฒŒ ๋ฉ๋‹ˆ๋‹ค. (์‹ค์ œ๋กœ ์„ธ์ฝฐ์ด์–ด ๋ฆฌํฌํŠธ์— ๋”ฐ๋ฅด๋ฉด ๊ธฐ์กด ๊ธฐ์—…๋“ค์— ๋น„ํ•ด AI๊ธฐ์—…๋“ค์˜ ๋ฆฌํ…์…˜์€ ์•„์ง ๋งค์šฐ ๋‚ฎ์Šต๋‹ˆ๋‹ค)

์ด๋Ÿฐ ํ˜„์ƒ์„ ์ €ํฌ๋Š” โ€˜์ž๋™์˜ ํ•จ์ •โ€™์ด๋ผ ๋ถ€๋ฆ…๋‹ˆ๋‹ค.

๊ทธ๋Ÿผ ๋‚˜๋จธ์ง€ 30์ ์€ ์–ด๋–ป๊ฒŒ ์ฑ„์›Œ์•ผ ํ• ๊นŒ์š”?
์—”์ง€๋‹ˆ์–ด๋ง๊ณผ UX๋ ˆ์ด์–ด๋กœ ๋งŽ์€ ๋ถ€๋ถ„์„ ํ•ด๊ฒฐํ•  ์ˆ˜ ์žˆ์—ˆ์Šต๋‹ˆ๋‹ค.

์ €ํฌ ์ œํ’ˆ์œผ๋กœ ์‚ฌ๋ก€๋ฅผ ๋“ค์–ด๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค.
๋ฆด๋ฆฌ์ŠคAI๋Š” ์˜์ƒ ์š”์•ฝ ์„œ๋น„์Šค๋ฅผ ์ œ๊ณตํ•˜๊ณ  ์žˆ๋Š”๋ฐ์š”,
์š”์•ฝ์ด ๊ฐ„ํ˜น ๋ถ€์ •ํ™•ํ•˜๊ฑฐ๋‚˜ ์ง€๋‚˜์น˜๊ฒŒ ์ถ•์•ฝ๋˜์–ด ์‹ ๋ขฐ๋„์— ์˜์‹ฌ์ด ์ƒ๊ธฐ๋Š” ๋ฌธ์ œ๊ฐ€ ์žˆ์—ˆ์Šต๋‹ˆ๋‹ค. ๊ธฐ์กด์˜ ๋‹ค๋ฅธ ์˜์ƒ ์š”์•ฝ ์„œ๋น„์Šค๋ฅผ ์ด์šฉํ•ด๋ดค์„ ๋•Œ, ์‹ ๋ขฐ๋„๊ฐ€ ์˜์‹ฌ๊ฐˆ ๋•Œ ์‚ฌ์šฉ์ž๊ฐ€ ์ทจํ•  ์ˆ˜ ์žˆ๋Š” ์•ก์…˜์ด ์—†์–ด ๊ฒฐ๊ตญ์—” ์˜์ƒ์„ ์ฒ˜์Œ๋ถ€ํ„ฐ ๋‹ค์‹œ ๋Œ๋ ค๋ณด๊ฒŒ ๋˜๋Š” ๋ถˆํŽธํ•จ์„ ๋А๊ผˆ์Šต๋‹ˆ๋‹ค.

๊ทธ๋ž˜์„œ ์ €ํฌ๋Š” 1) ์š”์•ฝ ๋ฌธ๋‹จ๋ณ„๋กœ ์š”์•ฝ์˜ ๊ทผ๊ฑฐ๊ฐ€ ๋˜๋Š” ์›๋ฌธ์Šคํฌ๋ฆฝํŠธ๋ฅผ ๋ฐ”๋กœ ์•„๋ž˜์— ๋ฐฐ์น˜ํ•˜๊ณ , 2) ์˜์ƒ ์žฌ์ƒ ํ”Œ๋ ˆ์ด์–ด๋ฅผ ์ขŒ์ธก์— ํ•ญ์ƒ ๊ณ ์ •์‹œ์ผœ ์š”์•ฝ์— ํ•ด๋‹นํ•˜๋Š” ๊ตฌ๊ฐ„์˜ ์˜์ƒ๊ณผ ์š”์•ฝ๋…ธํŠธ๋ฅผ ํ•จ๊ป˜ ๋ณผ ์ˆ˜ ์žˆ๋„๋ก ์‚ฌ์šฉ์ž ๊ฒฝํ—˜์„ ์„ค๊ณ„ํ–ˆ์Šต๋‹ˆ๋‹ค. 3) ์ตœ๊ทผ์—๋Š” ์ดํ•ด๊ฐ€ ์ž˜ ์•ˆ ๊ฐ€๋Š” ์šฉ์–ด๋ฅผ ํด๋ฆญํ•˜๋ฉด ์˜์ƒ์˜ ๋งฅ๋ฝ ์•ˆ์—์„œ ํ’€์ดํ•ด์ฃผ๋Š” ๊ธฐ๋Šฅ๋„ ์ถ”๊ฐ€ํ–ˆ์Šต๋‹ˆ๋‹ค.

(UX ๋ณด๋Ÿฌ๊ฐ€๊ธฐ : https://lnkd.in/grYf4ygu)

์ด๋Ÿฌํ•œ ์ œํ’ˆ ์ฒ ํ•™์„ ๊ฐ€์ง€๊ณ  ๊ณ„์† ์„œ๋น„์Šค๋ฅผ ๊ฐœ์„ ํ•˜๋‹ค๋ณด๋‹ˆ, D+1 Retention๋„ 30%๋Œ€์—์„œ ์‹œ์ž‘ํ•ด ํ˜„์žฌ 56%๋กœ ๊ณ„์† ์ฆ๊ฐ€ํ•˜๋Š” ์ค‘์ž…๋‹ˆ๋‹ค.

https://www.linkedin.com/posts/yein-kim-a33627219_ai-%EC%A0%9C%ED%92%88%EC%9D%98-ux%EB%8A%94-%EB%AC%B4%EC%97%87%EC%9D%B4-%EB%8B%AC%EB%9D%BC%EC%95%BC%ED%95%A0%EA%B9%8C-1-%EC%9E%90%EB%8F%99%EC%9D%98-%ED%95%A8%EC%A0%95-vrew-activity-7167747546352742401-VLTq?utm_source=share&utm_medium=member_desktop
You just keep grinding. 90%of the battle is just staying alive in business and having oxygen to keep trying things and then eventually if you are lucky things can really break in your favors.
โ€ข ๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ๊ฐ€ ํ”„๋ž‘์Šค ์ธ๊ณต์ง€๋Šฅ ์Šคํƒ€ํŠธ์—… ๋ฏธ์ŠคํŠธ๋ž„๊ณผ ํ˜‘์•ฝ์„ ๋งบ์Œ
โ€ข ์ด ํ˜‘์•ฝ์€ ๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ๊ฐ€ OpenAI ์ด์™ธ์˜ ๋ถ„์•ผ๋กœ ํ™œ๋™์„ ๋„“ํžˆ๋ ค๋Š” ์‹œ๋„์˜ ์ผํ™˜
โ€ข ๋ฏธ์ŠคํŠธ๋ž„์€ ํŒŒ๋ฆฌ ๊ธฐ๋ฐ˜์˜ ํšŒ์‚ฌ๋กœ, AI ๋ชจ๋ธ์„ ์‹œ์žฅ์— ์ถœ์‹œํ•˜๋Š” ๋ฐ ๋„์›€์„ ๋ฐ›๊ฒŒ ๋จ
ํ˜‘์•ฝ์˜ ๋‚ด์šฉ
โ€ข ๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ๋Š” ๋ฏธ์ŠคํŠธ๋ž„์— ํˆฌ์žํ•˜๋ฉฐ, ๊ธˆ์œต์ ์ธ ์„ธ๋ถ€ ์‚ฌํ•ญ์€ ๊ณต๊ฐœ๋˜์ง€ ์•Š์Œ
โ€ข ํ˜‘์•ฝ์—๋Š” ์œ ๋Ÿฝ ์ „์—ญ์˜ ์ •๋ถ€๋ฅผ ์œ„ํ•œ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜ ๊ฐœ๋ฐœ์„ ์œ„ํ•œ ์—ฐ๊ตฌ ๋ฐ ๊ฐœ๋ฐœ ํ˜‘๋ ฅ ํฌํ•จ
โ€ข ์ด AI ๋ชจ๋ธ๋“ค์€ ๊ณต๊ณต ๋ถ€๋ฌธ์˜ ํŠน์ • ์š”๊ตฌ ์‚ฌํ•ญ์„ ํ•ด๊ฒฐํ•˜๋Š” ๋ฐ ์‚ฌ์šฉ๋  ์˜ˆ์ •
๋ฏธ์ŠคํŠธ๋ž„์˜ ์œ„์น˜
โ€ข ๋ฏธ์ŠคํŠธ๋ž„์€ ๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ์˜ Azure ํด๋ผ์šฐ๋“œ ์ปดํ“จํŒ… ํ”Œ๋žซํผ์—์„œ ์ƒ์—…์  ์–ธ์–ด ๋ชจ๋ธ์„ ์ œ๊ณตํ•˜๋Š” ๋‘ ๋ฒˆ์งธ ํšŒ์‚ฌ
โ€ข ๋ฏธ์ŠคํŠธ๋ž„์˜ ๊ณต๋™ ์ฐฝ๋ฆฝ์ž ๊ฒธ CEO์ธ ์•„์„œ ๋ฉ˜์‰ฌ๋Š” ๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ์˜ ์‹ ๋ขฐ๊ฐ€ ์ค‘์š”ํ•œ ๋ฐœ์ „์ด๋ผ๊ณ  ์–ธ๊ธ‰
๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ์™€ OpenAI
โ€ข ๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ๋Š” ์ด๋ฏธ ์ƒŒํ”„๋ž€์‹œ์Šค์ฝ” ๊ธฐ๋ฐ˜์˜ OpenAI์— ์•ฝ 130์–ต ๋‹ฌ๋Ÿฌ๋ฅผ ํˆฌ์ž
โ€ข ๋ฏธ๊ตญ, EU, ์˜๊ตญ์˜ ๊ฒฝ์Ÿ ๊ฐ์‹œ ๊ธฐ๊ด€์—์„œ ์ด ํ˜‘๋ ฅ ๊ด€๊ณ„๋ฅผ ๊ฒ€ํ†  ์ค‘
โ€ข ๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ๋Š” OpenAI๋ฅผ ํ†ต์ œํ•˜์ง€ ์•Š์œผ๋ฉฐ, ๋‘ ํšŒ์‚ฌ๋Š” ์ค‘์š”ํ•œ ํŒŒํŠธ๋„ˆ๋ผ๊ณ  ๊ฐ•์กฐ
AI ๊ฒฝ์ œ์˜ ๋ถ€์ƒ
โ€ข ๊ตฌ๊ธ€๊ณผ ์•„๋งˆ์กด๊ณผ ๊ฐ™์€ ๋‹ค๋ฅธ ๋Œ€ํ˜• ๊ธฐ์ˆ  ํšŒ์‚ฌ๋“ค๋„ ์ƒ์„ฑ AI์— ๋Œ€ํ•œ ํˆฌ์ž๋ฅผ ํ™•๋Œ€
โ€ข ์ƒ์„ฑ AI๋Š” ํ…์ŠคํŠธ, ์ด๋ฏธ์ง€, ์ฝ”๋“œ๋ฅผ ๋ช‡ ์ดˆ ๋งŒ์— ์ƒ์„ฑํ•  ์ˆ˜ ์žˆ๋Š” ์†Œํ”„ํŠธ์›จ์–ด
โ€ข ์ด ๊ธฐ์ˆ ์€ ์ „ ์„ธ๊ณ„ ์‚ฐ์—…์„ ๋ณ€ํ™”์‹œํ‚ฌ ๊ฒƒ์œผ๋กœ ๋ถ„์„๊ฐ€๋“ค์€ ๋ฏฟ์Œ
๋ฏธ์ŠคํŠธ๋ž„์˜ ๊ธฐ์ˆ ๊ณผ ๊ฐ€์น˜
โ€ข ๋ฏธ์ŠคํŠธ๋ž„์€ ๋Œ€๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ์„ ๊ตฌ์ถ•ํ•˜๋ฉฐ, ์ด๋Š” ์ƒ์„ฑ AI ์ œํ’ˆ์„ ๊ตฌ๋™ํ•˜๋Š” ๊ธฐ์ˆ 
โ€ข 12์›”์— ์•ฝ 20์–ต ์œ ๋กœ์˜ ๊ฐ€์น˜ ํ‰๊ฐ€๋ฅผ ๋ฐ›์œผ๋ฉฐ ์•ฝ 4์–ต ์œ ๋กœ ๊ทœ๋ชจ์˜ ํŽ€๋”ฉ ๋ผ์šด๋“œ๋ฅผ ํ™•๋ณด
โ€ข ๋ฏธ์ŠคํŠธ๋ž„์˜ ๋ชจ๋ธ์€ ์˜คํ”ˆ ์†Œ์Šค๋กœ, ๊ธฐ์ˆ  ์„ธ๋ถ€ ์‚ฌํ•ญ์ด ๊ณต๊ฐœ์ ์œผ๋กœ ๊ณต๊ฐœ๋  ์˜ˆ์ •
AI ๋ชจ๋ธ ๊ฐœ๋ฐœ์˜ ์ธํ”„๋ผ
โ€ข ์ƒˆ๋กœ์šด AI ๋ชจ๋ธ์„ ํ›ˆ๋ จํ•˜๊ณ  ๊ฐœ๋ฐœํ•˜๋Š” ๋ฐ ํ•„์š”ํ•œ ์ธํ”„๋ผ๋Š” ๋ง‰๋Œ€ํ•œ ๋น„์šฉ์ด ๋“ฌ
โ€ข ๊ฒฝ์Ÿํ•  ์ˆ˜ ์žˆ๋Š” ํšŒ์‚ฌ๋Š” ๋ช‡ ์‹ญ ๊ฐœ์— ๋ถˆ๊ณผ
โ€ข ๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ๋Š” ๋…์ผ๊ณผ ์ŠคํŽ˜์ธ์— ์ƒˆ๋กœ์šด AI ๋ฐ์ดํ„ฐ ์„ผํ„ฐ์— 56์–ต ๋‹ฌ๋Ÿฌ๋ฅผ ํˆฌ์žํ•  ๊ณ„ํš ๋ฐœํ‘œ
AI ์นฉ ์„ค๊ณ„ ๋ฐ ๊ฐœ๋ฐœ์— ๋Œ€ํ•œ ํˆฌ์ž
โ€ข AI ์นฉ ์„ค๊ณ„ ๋ฐ ๊ฐœ๋ฐœ์— ๋” ๋งŽ์€ ํˆฌ์ž๊ฐ€ ์ด๋ฃจ์–ด์งˆ ๊ฒƒ์œผ๋กœ ์˜ˆ์ƒ
โ€ข ์—”๋น„๋””์•„๋Š” ์ด ๋ถ„์•ผ๋ฅผ ์„ ๋„ํ•˜๋ฉฐ, ์ตœ๊ทผ ๊ฐ€์น˜ ์ˆœ์œ„์—์„œ ์•„๋งˆ์กด๊ณผ ๊ตฌ๊ธ€ ๋ชจํšŒ์‚ฌ ์•ŒํŒŒ๋ฒณ์„ ์•ž์ง€๋ฆ„
โ€ข ๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ๋Š” ์ธํ…”๊ณผ AMD์™€ ๊ฐ™์€ ๊ฒฝ์Ÿ์‚ฌ์™€ ํ•จ๊ป˜ ์ž์ฒด ๋ฐ˜๋„์ฒด๋ฅผ ๊ฐœ๋ฐœํ•  ํˆฌ์ž ๊ณ„ํš
GNโบ์˜ ์˜๊ฒฌ
โ€ข ์ด ํ˜‘์•ฝ์€ ๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ๊ฐ€ AI ๋ถ„์•ผ์—์„œ ์ž์‚ฌ์˜ ์˜ํ–ฅ๋ ฅ์„ ํ™•๋Œ€ํ•˜๊ณ ์ž ํ•˜๋Š” ์ „๋žต์  ์›€์ง์ž„์„ ๋ณด์—ฌ์คŒ. OpenAI์™€์˜ ํ˜‘๋ ฅ ์™ธ์—๋„ ๋‹ค์–‘ํ•œ AI ๊ธฐ์ˆ ์„ ํ™•๋ณดํ•˜๋ ค๋Š” ์˜๋„๊ฐ€ ์—ฟ๋ณด์ž„
โ€ข ๋ฏธ์ŠคํŠธ๋ž„๊ณผ์˜ ํŒŒํŠธ๋„ˆ์‹ญ์€ ๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ๊ฐ€ ์˜คํ”ˆ ์†Œ์Šค AI ๋ชจ๋ธ์— ๋Œ€ํ•œ ์ง€์›์„ ๊ณ„์†ํ•˜๊ฒ ๋‹ค๋Š” ์˜์ง€๋ฅผ ๋‚˜ํƒ€๋ƒ„. ์ด๋Š” ๊ฐœ๋ฐœ์ž ์ปค๋ฎค๋‹ˆํ‹ฐ์— ๋” ๋งŽ์€ ํ˜‘๋ ฅ๊ณผ ํ˜์‹ ์˜ ๊ธฐํšŒ๋ฅผ ์ œ๊ณตํ•  ์ˆ˜ ์žˆ์Œ
โ€ข AI ๊ธฐ์ˆ ์˜ ๋น ๋ฅธ ๋ฐœ์ „๊ณผ ์ƒ์—…ํ™”๋Š” ๋ฐ์ดํ„ฐ ํ”„๋ผ์ด๋ฒ„์‹œ, ์•Œ๊ณ ๋ฆฌ์ฆ˜ ํˆฌ๋ช…์„ฑ, ๊ทธ๋ฆฌ๊ณ  ๊ธฐ์ˆ ์˜ ์œค๋ฆฌ์  ์‚ฌ์šฉ์— ๋Œ€ํ•œ ๋…ผ์˜๋ฅผ ๋”์šฑ ์ค‘์š”ํ•˜๊ฒŒ ๋งŒ๋“ฆ. ์ด๋Ÿฌํ•œ ์ด์Šˆ๋“ค์€ AI ๊ธฐ์ˆ ์„ ๋„์ž…ํ•  ๋•Œ ๋ฐ˜๋“œ์‹œ ๊ณ ๋ คํ•ด์•ผ ํ•  ์‚ฌํ•ญ์ž„
โ€ข ๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ๊ฐ€ AI ์นฉ ์„ค๊ณ„ ๋ฐ ๊ฐœ๋ฐœ์— ํˆฌ์žํ•  ๊ณ„ํš์„ ๋ฐํžŒ ๊ฒƒ์€, AI ๊ธฐ์ˆ ์˜ ํ•ต์‹ฌ ๊ตฌ์„ฑ ์š”์†Œ์ธ ํ•˜๋“œ์›จ์–ด์— ๋Œ€ํ•œ ํ†ต์ œ๋ฅผ ๊ฐ•ํ™”ํ•˜๋ ค๋Š” ์‹œ๋„๋กœ ๋ณผ ์ˆ˜ ์žˆ์Œ. ์ด๋Š” ์žฅ๊ธฐ์ ์œผ๋กœ AI ์‹œ์žฅ์—์„œ์˜ ๊ฒฝ์Ÿ๋ ฅ์„ ๋†’์ด๋Š” ๋ฐ ๊ธฐ์—ฌํ•  ์ˆ˜ ์žˆ์Œ
โ€ข AI ๋ชจ๋ธ์„ ๊ฐœ๋ฐœํ•˜๊ณ  ํ›ˆ๋ จํ•˜๋Š” ๋ฐ ํ•„์š”ํ•œ ์ธํ”„๋ผ์˜ ๋น„์šฉ๊ณผ ๋ณต์žก์„ฑ์„ ๊ณ ๋ คํ•  ๋•Œ, ํด๋ผ์šฐ๋“œ ์„œ๋น„์Šค ์ œ๊ณต์—…์ฒด์™€์˜ ํ˜‘๋ ฅ์€ ์ค‘์†Œ๊ธฐ์—…์ด๋‚˜ ์Šคํƒ€ํŠธ์—…์—๊ฒŒ ์ค‘์š”ํ•œ ์ž์›์ด ๋  ์ˆ˜ ์žˆ์Œ. ์ด๋Ÿฌํ•œ ํ˜‘๋ ฅ์€ ๊ธฐ์ˆ  ํ˜์‹ ๊ณผ ์‹œ์žฅ ๋‹ค์–‘์„ฑ์„ ์ด‰์ง„ํ•˜๋Š” ๋ฐ ์ค‘์š”ํ•œ ์—ญํ• ์„ ํ•  ๊ฒƒ์ž„

https://www.ft.com/content/cd6eb51a-3276-450f-87fd-97e8410db9eb
Continuous Learning_Startup & Investment
OpenAI also is aiming to build a simulator which mimic the reality without gaps.
Modeling the world for action by generating pixels is as wasteful and doomed to failure.

If your goal is to train a world model for recognition or planning, using pixel-level prediction is a terrible idea," writes LeCun.
First of all, really appreciate the team for sharing helpful insights and design decisions โ€“ Sora is incredible and is set to transform the video generation community.

What we have learned so far:
- Architecture: Sora is built on our diffusion transformer (DiT) model (published in ICCV 2023) โ€” it's a diffusion model with a transformer backbone, in short:
DiT = [VAE encoder + ViT + DDPM + VAE decoder].
According to the report, it seems there are not much additional bells and whistles.

- "Video compressor network": Looks like it's just a VAE but trained on raw video data. Tokenization probably plays a significant role in getting good temporal consistency. By the way, VAE is a ConvNet, so DiT technically is a hybrid model ;)

When Bill and I were working on the DiT project, instead of creating novelty (see my last tweet๐Ÿคทโ€โ™‚๏ธ), we prioritized two aspects: simplicity and scalability. These priorities offer more than just conceptual advantages.

- Simplicity means flexibility. The cool thing about vanilla ViT that people often miss is how it makes your model way more flexible when it comes to working with input data. For example, in masked autoencoder (MAE), ViT helped us to just process the visible patches and ignore the masked ones. And similarly, Sora "can control the size of generated videos by arranging randomly-initialized patches in an appropriately-sized grid." UNet does not directly offer this flexibility.
๐Ÿ‘€Speculation: Sora might also use Patch nโ€™ Pack (NaViT) from Google, to make DiT adaptable to variable resolutions/durations/aspect ratios.

- Scalability is the core theme of the DiT paper. First, an optimized DiT runs much faster than UNet in terms of wall-clock time per Flop. More importantly, Sora demonstrated that the DiT scaling law applies not just to images but now to videos as well -- Sora replicates the visual scaling behavior observed in DiT.
๐Ÿ‘€Speculation: In the Sora report, the quality for the first video is quite bad, I suspect it is using a base model size. A back-of-the-envelope calculation: DiT XL/2 is 5X GFLOPs of the B/2 model, so the final 16X compute model is probably 3X DiT-XL model size, which means Sora might have ~3B parameters โ€“ if true, this is not an unreasonable model size . It could suggest that training the Sora model might not require as many GPUs as one would anticipate โ€“ I would expect very fast iterations going forward.

The key takeaway is from the "Emerging simulation capabilities" section. Before Sora, it was unclear if long form consistency could emerge on its own or if it required complex subject-driven generation pipelines or even physics simulators. OpenAI has shown that, though not perfect, these behaviors can be achieved with end-to-end training. Yet, two essential points have not been discussed.

1. Training Data: No talk about training source and construction at all, which might just imply data is likely the most critical factor for Sora's success.
๐Ÿ‘€Speculations: There's already much speculation about data from game engines. I also anticipate the inclusion of movies, documentaries, cinematic long takes, etc. Quality really matters. Super curious where Sora got this data from (surely not YouTube, right?).

2. (Auto-regressive) Long Video Generation: a significant breakthrough in Sora is the ability to generate very long videos. The difference between producing a 2-second video and a 1-minute video is monumental.
In Sora, this is probably achieved through joint frame prediction that allows auto-regressive sampling, yet a major challenge is how to address error accumulation and maintain quality/consistency through time. A very long (and bi-directional) context for conditioning? Or could scaling up simply lessen the issue? These technical details can be super important and hopefully will be demystified in the future

https://x.com/gabor/status/1758295719788822866?s=46&t=h5Byg6Wosg8MJb4pbPSDow

DiT shines in Sora. Our team at NYU has recently released a new DiT model, called SiT.
์ง€๋‚œ ์›”์š”์ผ(11์›” 27์ผ), ์•Œํ† ์Šค ์˜คํ”ผ์Šค์—์„œ ์Šคํƒ€ํŠธ์—… ๋ฆฌ๋”๋“ค๊ณผ์˜ ๋ชจ์ž„์„ ์ฃผ์ตœํ–ˆ์Šต๋‹ˆ๋‹ค. ๊ณต๊ฐ„์„ ๋งˆ๋ จํ•ด์ฃผ์‹  ํ˜ธ์ŠคํŒ… ์˜ค์˜ํƒ ๋‹˜๊ณผ ์•Œํ† ์Šค ๊ด€๊ณ„์ž๋ถ„๋“ค๊ป˜ ๊ฐ์‚ฌ์˜ ๋ง์”€์„ ์ „ํ•ฉ๋‹ˆ๋‹ค.

์ œํ•œ๋œ ๊ณต๊ฐ„๊ณผ ์‹œ๊ฐ„์œผ๋กœ ์‹ ์ฒญํ•ด์ฃผ์‹  ๋ชจ๋“  ๋ถ„๋“ค๊ณผ ํ•จ๊ป˜ํ•˜์ง€ ๋ชปํ•œ ์ ์€ ์•„์‰ฌ์› ์Šต๋‹ˆ๋‹ค. CTO, ๊ฐœ๋ฐœ ๋ฆฌ๋“œ, CEO๋ถ„๋“ค๊ณผ '๋™๊ธฐ๋ถ€์—ฌ', 'ํŒ€์˜ ๋ถˆ์™„์ „์„ฑ(ํŠนํžˆ ์‹œ๋‹ˆ์–ด์˜ ๋ถ€์žฌ)', ๊ทธ๋ฆฌ๊ณ  'ํƒ€ ํŒ€๊ณผ์˜ ์†Œํ†ต ๋ฌธ์ œ'๋ผ๋Š” ์ฃผ์ œ๋กœ ์‹œ์ž‘ํ•œ ํ† ๋ก ์€ ํฅ๋ฏธ์ง„์ง„ํ–ˆ์Šต๋‹ˆ๋‹ค. ๋ชจ๋“  ๋ถ„๋“ค์ด ์—ด์ •์ ์œผ๋กœ ์ฐธ์—ฌํ•ด์ฃผ์‹œ๊ณ  ๋งŽ์€ ์ธ์‚ฌ์ดํŠธ์™€ ๊ณ ๋ฏผ์„ ํ•จ๊ป˜ ๋‚˜๋ˆ  ์ฃผ์…”์„œ ๋‹ค์‹œ ํ•œ๋ฒˆ ๋„ˆ๋ฌด ๊ฐ์‚ฌ๋“œ๋ฆฝ๋‹ˆ๋‹ค.

์ €๋Š” 2๋…„๊ฐ„์˜ ๋ธ”๋ก์ฒด์ธ ์˜คํ”ˆ์†Œ์Šค ํ™œ๋™๊ณผ ์•ฝ 5๋…„๊ฐ„ ํ•ด์น˜๋žฉ์Šค์—์„œ ๊ฒฝํ—˜์„ ๋ฐ”ํƒ•์œผ๋กœ Engineering Manager ๊ฒฝํ—˜์„ ๊ณต์œ ํ–ˆ์Šต๋‹ˆ๋‹ค. ๋ชจ์ž„์—์„œ ๊ณต์œ ํ•œ ๋‚ด์šฉ์„ ์ €์ฒ˜๋Ÿผ ํŒ€์„ ์œ„ํ•ด ๋ฐค๋‚ฎ์œผ๋กœ ๊ณ ๋ฏผํ•˜์‹œ๋Š” ๋ถ„๋“ค๊ป˜ ์กฐ๊ธˆ์ด๋ผ๋„ ๋„์›€์ด ๋˜๊ณ ์ž ์งง๊ฒŒ ๊ณต์œ ๋“œ๋ฆฝ๋‹ˆ๋‹ค. ๐Ÿ”

1. ์ €๋Š” Manager๋ฅผ ํŒ€(์‚ฌ๋žŒ)์— ์ง‘์ค‘ํ•˜์—ฌ ์กฐ์ง์—์„œ ์˜ํ–ฅ๋ ฅ์„ ๋ฐœํœ˜ํ•˜๋Š” ์‚ฌ๋žŒ์œผ๋กœ ์ •์˜ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.

2. IC(๊ฐœ๋ณ„ ๊ธฐ์—ฌ์ž)์—์„œ Manager๋กœ ์ „ํ™˜ํ•˜๋ฉด์„œ ์ •์ฒด์„ฑ, ๊ด€๊ณ„, ์˜ํ–ฅ๋ ฅ์˜ ์ „ํ™˜์ด ์ผ์–ด๋‚ฌ์Šต๋‹ˆ๋‹ค.
- ์ •์ฒด์„ฑ์˜ ์ „ํ™˜: '์ข‹์€ ์ œํ’ˆ์„ ๋งŒ๋“œ๋Š” ๊ฐœ๋ฐœ์ž'์—์„œ '๋‚ด์žฌ์  ๋™๊ธฐ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ํŒ€์›๋“ค์˜ ์„ฑ ์žฅ์„ ์ด‰์ง„ํ•˜๋ฉฐ, ์ง€์†๊ฐ€๋Šฅํ•˜๊ณ  ๊ฐ€์น˜์žˆ๋Š” ๊ฒฐ๊ณผ๋ฅผ ๋‹ฌ์„ฑํ•˜๋Š” ํŒ€์„ ๋งŒ๋“ค๊ณ  ์ด๋Œ์–ด ๊ฐ€๋Š” ๋ฆฌ๋”๋กœ ์ „ํ™˜'
- ๊ด€๊ณ„์˜ ์ „ํ™˜: ํ•ต์‹ฌ ์ดํ•ด๊ด€๊ณ„์ž๋“ค์ด '์—”์ง€๋‹ˆ์–ด ํŒ€์›๋“ค'์—์„œ 'CEO, Head of BD, Head of TA, ์ฃผ์š” ์—”์ง€๋‹ˆ์–ด ํŒ€์›'์œผ๋กœ ์ „ํ™˜
- ์˜ํ–ฅ๋ ฅ์˜ ์ „ํ™˜: '์ œํ’ˆ ๊ฐœ๋ฐœ'์—์„œ '๊ตฌ์„ฑ์›๋“ค์˜ ์„ฑ๊ณผ & ์—ญ๋Ÿ‰ ์„ฑ์žฅ ๋ฐ ํ‰๊ฐ€, ํ”ผ๋“œ๋ฐฑ, ์ฑ„์šฉ, ํŒ€์˜ ๋ฌธํ™”์™€ ๋ฐฉํ–ฅ์„ฑ ๊ตฌ์ถ•'์œผ๋กœ ์ „ํ™˜

3. Manager๋กœ ํŒ€์˜ ๋ฏธ์…˜์„ ๋‹ฌ์„ฑํ•˜๊ธฐ ์œ„ํ•œ ํŒ€์˜ Being(์กด์žฌ) ์„ค์ •ํ–ˆ์Šต๋‹ˆ๋‹ค.
- ์‹œ์žฅ ๋ณ€ํ™”์— ์‹ ์†ํ•˜๊ฒŒ ๋Œ€์‘ํ•˜๊ณ  ์ง€์†์ ์œผ๋กœ ๋น ๋ฅด๊ฒŒ ๊ฐ€์น˜๋ฅผ ์ฐฝ์ถœํ•˜๋Š” ์—”์ง€๋‹ˆ์–ด๋ง ํŒ€
- ๋‹จ์ˆœ ๊ฐœ๋ฐœ์„ ๋„˜์–ด ํ†ตํ•ฉ๋œ ์ฑ…์ž„๊ฐ์„ ๊ฐ€์ง„ ์—”์ง€๋‹ˆ์–ด๋ง ํŒ€

4. Top Down, Bottom Up ์ „๋žต์„ ํ™œ์šฉํ•˜์—ฌ ๋ชฉํ‘œ๋ฅผ ๋‹ฌ์„ฑํ•˜๊ธฐ ์œ„ํ•ด ๋…ธ๋ ฅํ–ˆ์Šต๋‹ˆ๋‹ค.
Top Down ์ „๋žต (Manager๋กœ ๋ฐฉํ–ฅ์„ฑ ์ œ์‹œ)
- ๊ฐœ๋ฐœ ์ž‘์—…์˜ ๋‹จ์œ„์™€ ์ฑ…์ž„ ๋ฒ”์œ„ ์กฐ์ •
- ๊ธฐ๋Šฅ๋ณ„ ๋ฐฐํฌ ์ „๋žต & E2E ์ž๋™ํ™” ํ…Œ์ŠคํŠธ ๋„์ž…
- ์—ญ๋Ÿ‰ ํ”„๋ ˆ์ž„์›Œํฌ ์ •์˜
- ์ƒ์‚ฐ์„ฑ ์ง€ํ‘œ ๋„์ž…
Bottom Up ์ „๋žต (ํŒ€์›์—๊ฒŒ ์ดˆ์ ์„ ๋งž์ถฐ ๋‚ด์žฌ์  ๋™๊ธฐ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ํŒ€ ๋ฐฉํ–ฅ์„ฑ์— Align)
- 2์ฃผ ๋‹จ์œ„์˜ 1on1(๋ฐฉํ–ฅ์„ฑ Align & ์„ฑ์žฅ์— ์ดˆ์ )
- ์—ญ๋Ÿ‰ ํ‰๊ฐ€

5. ๋Œ€๋ถ€๋ถ„์˜ ์–ด๋ ค์›€์€ Bottom Up ์ „๋žต์„ ์‹คํ–‰ํ•˜๋Š” ๋ฐ์—์„œ ์ผ์–ด๋‚ฌ์Šต๋‹ˆ๋‹ค.

6. ์•„์ง ๋งŽ์ด ๋ถ€์กฑํ•˜๋‹ค๊ณ  ์ƒ๊ฐํ•˜์ง€๋งŒ, ํŒ€์˜ ์ƒ์‚ฐ์„ฑ์„ ํฌ๊ฒŒ ํ–ฅ์ƒ ์‹œ์ผฐ๊ณ , ํ†ตํ•ฉ๋œ ์ฑ…์ž„๊ฐ์„ ๋А๋ผ๋Š” ๋ฌธํ™”๊ฐ€ ์ž๋ฆฌ ์žก์•˜์œผ๋ฉฐ, ๊ธฐ๋Œ€ํ•˜๋Š” ์—ญ๋Ÿ‰์— ๋Œ€ํ•œ ํ•ฉ์˜๋ฅผ ํ†ตํ•ด ๊ตฌ์„ฑ์›๋“ค์˜ ์„ฑ์žฅ์ด ์ผ์–ด๋‚ฌ์Šต๋‹ˆ๋‹ค.

7. ์ด๋Ÿฐ ๊ณผ์ • ์†์—์„œ ์ œ๊ฐ€ ๋ฐฐ์šด์ ์€ ๋‹ค์Œ๊ณผ ๊ฐ™์Šต๋‹ˆ๋‹ค(๋‚ด์šฉ์€ ๋ป”ํ•˜์ง€๋งŒ ์ž˜ ํ•˜๋Š”๊ฒƒ์€ ์–ด๋ ต๋‹ค๊ณ  ์ƒ๊ฐํ•ฉ๋‹ˆ๋‹ค).
- Bottom Up ์ „๋žต์˜ ์ค‘์š”์„ฑ ์ธ์‹: ์•„๋ฌด๋ฆฌ ๋›ฐ์–ด๋‚œ ์ „๋žต์ด๋”๋ผ๋„ ๊ตฌ์„ฑ์›๋“ค์ด ์ œ๋Œ€๋กœ ์‹คํ–‰ํ•ด์ฃผ์ง€ ์•Š์œผ๋ฉด ์‹คํŒจํ•จ
- ๊ตฌ์„ฑ์›๋“ค์˜ ๋ฌธ์ œ์— ๋Œ€ํ•œ ๊นŠ์€ ์ดํ•ด: ๊ฒฝ์ฒญ๊ณผ ์งˆ๋ฌธ์„ ํ†ตํ•ด ๊ตฌ์„ฑ์›๋“ค์˜ ๋ฌธ์ œ์— ๋Œ€ํ•œ ๋ณธ์งˆ์  ์ ‘๊ทผ ๋ฐ ์ดํ•ด๊ฐ€ ์ค‘์š”ํ•จ
- ๋‚˜๋ณด๋‹ค ์—ญ๋Ÿ‰์ด ๋›ฐ์–ด๋‚œ IC์™€ ์ผํ•˜๋Š” ๋ฒ•: ๊ทธ๋“ค์ด ๋ฌธ์ œ์— ์ง‘์ค‘ํ• ๋•Œ ํŒ€์˜ ๊ด€์ ์—์„œ ๋ฐฉํ–ฅ์„ฑ์„ ์ œ์‹œํ•˜์—ฌ ํŒ€ ์„ฑ๊ณผ๋ฅผ ๋†’์ผ ์ˆ˜ ์žˆ์Œ
- ๊ตฌ์„ฑ์›์— ๋Œ€ํ•œ ๋ช…ํ™•ํ•œ ๊ธฐ๋Œ€์น˜ ์„ค์ •: ๊ตฌ์„ฑ์›์˜ ์„ฑ๊ณผ๋‚˜ ์—ญ๋Ÿ‰์ด ์•„์‰ฝ๋‹ค๋ฉด ๋ช…ํ™•ํ•˜๊ฒŒ ๊ธฐ๋Œ€ํ•˜๋Š” ๋ฐ”๋ฅผ ์ „๋‹ฌํ–ˆ๋Š”์ง€๋ฅผ ๊ณ ๋ฏผํ•ด์•ผํ•จ
๐Ÿ‘1
์€์‚ฌ๋‹˜๊ณผ์˜ ๋Œ€ํ™” ์ค‘ ๊ธฐ์–ต์— ๋‚จ๋Š” ๊ตฌ์ ˆ๋“ค.

๊ฒฐ๊ตญ ์ฐฝ์—…์ž์˜ ๊ฐ€์น˜๊ด€, ์„ธ๊ณ„๊ด€์ด ๊ทธ ์กฐ์ง์˜ DNA์ด๊ณ  ์ผํ•˜๋Š” ๋ฐฉ์‹์ด๋‹ค.

์ฐฝ์—…์ž์˜ ๋‚˜๋‹ค์›€์ด ์กฐ์ง์˜ ๋‚˜๋‹ค์›€์„ ๋งŒ๋“ ๋‹ค.

์ข‹์•„ํ•˜์ง€๋งŒ, ๊ณ ๊ฐ์„ ๋งŒ์กฑ์‹œํ‚ฌ๋งŒํผ ์ง€๋…ํ•˜๊ฒŒ ์ข‹์•„ํ•˜๋Š” ์ผ์„ ์ฐฝ์—…ํ•ด์•ผํ•œ๋‹ค.

์งง๊ณ  ๋น ๋ฅธ ์„ฑ๊ณต์ด ์•„๋‹ˆ๋ผ ๋‚˜๋‹ค์šด ์—ฌ์ •์„ ๊ฐ€๋Š” ๊ฒƒ์ด ์ค‘์š”ํ•˜๋‹ค.

๋„ˆ๋ฌด ์งง์€ ์ˆœ๊ฐ„์— ๋งŽ์€ ๊ฒƒ์„ ์ด๋ค˜๋‹ค๊ณ  ์ฐฉ๊ฐํ•˜๋ฉด ๋Š™๊ณ  ๊ณ ์ธ๋‹ค.

์ฃผ๋ณ€์„ ๋ณด๊ณ  ๋ฐฐ์šฐ๋ฉด ์„ฑ์žฅํ•  ์ˆ˜ ์žˆ๊ณ , ๋น„๊ตํ•˜๋ฉด ๋ถˆํ–‰ํ•ด์ง„๋‹ค.

๋˜๋Š”์ง€ ์•ˆ๋˜๋Š”์ง€๋Š” ํ•ด๋ณด๊ธฐ ์ „๊นŒ์ง„ ๋ชจ๋ฅธ๋‹ค. ๊ฐ€๋Šฅํ•œ ๋งŽ์€ ๊ฒƒ๋“ค์„ ํ•ด๋ณด๊ณ  ๋‚˜๋ฉด ๊ทธ ์ผ์ด ์ž˜ ์•ˆ๋˜๋”๋ผ๋„ ์ƒˆ๋กœ์šด ์„ธ์ƒ๊ณผ ๊ธฐํšŒ๊ฐ€ ๋ณด์ธ๋‹ค.

๋ˆ ๋งŽ์ด ๋ฒŒ์–ด์„œ ๋‚จ ์ข‹์€ ์ผ ๋งŽ์ด ํ•˜์ž.

ํŠน์ • ๋ชฉ์ ์„ ์„ฑ์ทจํ•˜๋Š” ์‚ถ์ด ์•„๋‹ˆ๋ผ ๊พธ์ค€ํžˆ ๋ฐฐ์šฐ๊ณ  ์„ฑ์žฅํ•˜๊ณ  ๊ฐ์‚ฌํ•˜๋Š” ๊ณผ์ •์ด ์ข‹๋‹ค.
โค2
Many people learn about the tools just enough to get the job done. I prefer to dive deeper; understanding my tools in detail makes my work much more fun.

Not sure if it's good or bad. Just more fun! Perhaps thatโ€™s what truly matters in the end.
๐Ÿ”ฅ2
A rock pile ceases to be a rock pile the moment a single man contemplates it, bearing within him the image of a cathedral.

ํ•œ ์‚ฌ๋žŒ์ด๋ผ๋„ ํฐ ์„ฑ๋‹น์˜ ์ด๋ฏธ์ง€๋ฅผ ํ’ˆ๊ณ  ๋Œ๋ฌด๋”๊ธฐ๋ฅผ ๋ณธ๋‹ค๋ฉด, ๊ทธ ์ˆœ๊ฐ„ ๋” ์ด์ƒ ๊ทธ๊ฒƒ์€ ๋Œ๋ฌด๋”๊ธฐ๊ฐ€ ์•„๋‹ˆ๋‹ค.
๐Ÿ’กWinning Instincts vs. Winning Ideas: Your instinct is correct 95% of the time; your ideas - 25% of the time. Trust your instincts, not your ideas.
๐Ÿ’ช Small Ideas - Big Ambitions: Start small with big ambitions.
๐Ÿงฉ Proven, Better, New: Each product is an aggregation of individual features where some features are โ€˜provenโ€™, other features are โ€˜betterโ€™ than the existing ones, and there are โ€˜newโ€™ features that did not exist before. It is about making the product, as a whole, better. Otherwise (reciting Steve Jobs), by trying to make every feature 10% better, you may make the product 50% worse.
๐Ÿ†• All New Fails: Assume new fails, and you can find a way to work around it.
๐Ÿก Death By a Thousand Compromises: If you make too many compromises as a founder, you may end up building a company you donโ€™t want to work at. Build a house you want to live in. Decide how ambitious you really are. No judgment, just intellectual honesty. Identify things that matter to you, lines you would not cross, trade-offs you would be willing to make. โ€œIโ€™m a team player; if Iโ€™m running the team, I have to be CEOโ€
๐ŸŽˆ Taste & creativity: You have to constantly increase your bar, develop your taste, make it better. Care at a pixel level. Keep an eye out for true signals.
โš™๏ธ Lessons Learned: Canโ€™t be naรฏve to a new space; have to learn on the go. Proven, better, new โ€“ copy what works well. Allow yourself to be too early โ€“ it may be hard but it has to be done to discover a new planet. Understand what your ego is and how to avoid pitfalls that come with it.
โœจ Happiness: Being and feeling useful to a community you care about.

Mark Pincus Founder and CEO of Zynga
https://www.klarna.com/international/press/klarna-ai-assistant-handles-two-thirds-of-customer-service-chats-in-its-first-month/

โ€ข The AI assistant has had 2.3 million conversations, two-thirds of Klarnaโ€™s customer service chats
โ€ข It is doing the equivalent work of 700 full-time agents
โ€ข It is on par with human agents in regard to customer satisfaction score
โ€ข It is more accurate in errand resolution, leading to a 25% drop in repeat inquiries
โ€ข Customers now resolve their errands in less than 2 mins compared to 11 mins previously
โ€ข Itโ€™s available in 23 markets, 24/7 and communicates in more than 35 languages
โ€ข Itโ€™s estimated to drive a $40 million USD in profit improvement to Klarna in 2024