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
πŸ“’ Introducing Voyage AI

Founded by a talented team of leading AI researchers and me πŸš€πŸš€.

We build state-of-the-art embedding models (e.g., better than OpenAI 😜).

We also offer custom models that deliver 🎯+10-20% accuracy gain in your LLM products. 🧡

We think embeddings are an under-loved, but incredibly important, part of everyone’s retrieval stack. So we set out to build just better embeddings.

Voyage embeddings are SOTA on MTEB and 9 held-out benchmarks covering industry domains. Learn more here: https://lnkd.in/gk59ZE-T

We partnered with @LangChainAI to help enhance their official chatbot, live at chat.langchain.com. Both our base and fine-tuned embeddings improve both the retrieval and response quality, and the latter is deployed!

Check out https://lnkd.in/gZ2aH4KU for more details.

Try Voyage embeddings for free – your first 5000 documents are on us!

Create an account at voyageai.com and check out our API docs at https://docs.voyageai.com .
To fine-tune your own custom model, email βœ‰οΈ contact@voyageai.com for early access!
🀩2
Long 베이비뢀머 vs Short MZ

BofA λŠ” 졜근 λ³΄κ³ μ„œμ—μ„œ Long 베이비뢀머, Short λ°€λ ˆλ‹ˆμ–Ό (MZ) 투자 아이디어λ₯Ό μ œμ‹œν•©λ‹ˆλ‹€.

μ§€κΈˆμ˜ 고금리, κ³ λ¬Όκ°€ ν™˜κ²½μ—μ„œ λ―Έκ΅­λ‚΄ μ†ŒλΉ„ μ§€μΆœμ˜ μ£Όλ„κΆŒμ€ 65μ„Έ μ΄μƒμ˜ 고령측이 μ™„λ²½νžˆ μ₯κ³  μžˆμŠ΅λ‹ˆλ‹€.

ν•œκ΅­μ˜ 베이비뢀머 μ„ΈλŒ€μΈ 저희 λΆ€λͺ¨λ‹˜λ“€μ„ 보더라도 λŒ€λΆ€λΆ„ μžκ°€λ‘œ 1 주택 이상을 보유 μ€‘μž…λ‹ˆλ‹€. λŒ€μΆœ 거의 μ—†μŠ΅λ‹ˆλ‹€. μžμ‚° κ°€κ²©μ˜ μΈν”Œλ ˆμ† Wealth effect λ¬΄μ‹œ λͺ»ν•©λ‹ˆλ‹€.

반면, MZ μ„ΈλŒ€λŠ” 주둜 μ“°κ³  μ£½μžλŠ” 세계관, 졜근 λͺ‡λ…„κ°„ 뢀동산 영끌쑱, 코인/주식 빚투쑱의 ν•©λ₯˜λ‘œ 뢀채 뢀담이 μƒλ‹Ήν•©λ‹ˆλ‹€. 숨만 쉬어도 이자의 Snowball effect λ₯Ό κ²½ν—˜ν•˜κ³  있죠.

λ”μš±μ΄ λ² μ΄λΉ„λΆ€λ¨ΈλŠ” 돈이 λ§Žλ“  적든 μ—¬μ „νžˆ 저좕에 λŒ€ν•œ 믿음이 μžˆμŠ΅λ‹ˆλ‹€. 미ꡭ도 Cash return 이 5% λ₯Ό μƒνšŒν•©λ‹ˆλ‹€. Cash is king 인 μž‘κΈˆμ˜ 상황은 μ„ΈλŒ€κ°„ 격차λ₯Ό 되돌리기 νž˜λ“€ μ •λ„λ‘œ 벌리고 μžˆμŠ΅λ‹ˆλ‹€.

μŠΉμžμ™€ νŒ¨μžκ°€ λͺ…ν™•νžˆ κ°ˆλ¦¬λŠ” κ²Œμž„μ΄λΌλ©΄, 섹터와 μ‚°μ—…λ‚΄ 투자 μ•„μ΄λ””μ–΄λ‘œ ν™œμš©ν•΄ 볼만 ν•©λ‹ˆλ‹€.
51% of Zendesk's revenue is international
47% of Cloudflare’s revenue is international
46% of HubSpot’s revenue is international
45% of MongoDB's revenue is international
38% of Asana’s revenue is international
34% of Twilio’s revenue is international
22% of Okta’s revenue is international

Go Global as soon as you have customers there

Pull ahead
β€œμ°½μ—…κ°€μ˜ 길에 λ“€μ–΄μ„œλ©΄ νšŒμ‚¬κ°€ μ—¬λŸ¬λΆ„μ˜ 가쑱보닀 μ€‘μš”ν•΄μ Έμš”. κ°€μ‘±λ“€μ˜ λŒ€μ†Œμ‚¬λ₯Ό μ±™κΈΈ 수 μ—†κ³ , μžλ…€λ“€μ—κ²Œ κ΅Ώλ‚˜μž‡ ν‚€μŠ€λ₯Ό ν•  수 μ—†κ²Œ 될 κ±°μ˜ˆμš”. 또 ν˜„κΈˆ μžμ‚°μ΄ ν•„μš”ν•΄ 쒋은 μ§‘κ³Ό 쒋은 μ°¨λŠ” 포기해야할 κ±°κ³ μš”. μΉœκ΅¬λ‚˜ 지인듀은 μ—¬λŸ¬λΆ„μ„ μ΄ν•΄ν•˜μ§€ λͺ»ν•˜κ³ , μΈμƒμ—μ„œ 길을 μžƒκ³  λ°©ν™©ν•œλ‹€κ³  생각할 κ±°μ˜ˆμš”. λΆ€λ”” 방황이 μ§§μ•„μ•Όν•  텐데 ν•˜κ² μ£ .”
β€œνŒ€μ›λ“€μ˜ 월급을 쀄 수 μ—†λŠ” μ‹œκΈ°κ°€ 올 κ±°μ˜ˆμš”. 이 μ—¬μ •μ—μ„œ λˆ„κ΅°κ°€λŠ” λ‹Ήμ‹ (μ°½μ—…κ°€)을 κ³ μ†Œν•  κ±°μ˜ˆμš”. μ—¬λŸ¬λΆ„μ„ κ³ μ†Œν•˜λŠ” μ‚¬λžŒμ΄ 곡동 μ°½μ—…μžμΌ μˆ˜λ„ 있고 투자자일 μˆ˜λ„ μžˆμ–΄μš”. 또 자주 틀릴 것이고, 이 λ¬Έμ œλ“€μ„ νŒ€μ›λ“€μ΄ λ‹€ 보게 될 κ±°μ˜ˆμš”. μͺ½νŒ”λ¦° 단계λ₯Ό λ„˜μ–΄μ„œλ©΄ λ‚΄κ°€ ν•˜λŠ” 게 λ§žλ‚˜, (λŒ€ν‘œμ§μ—μ„œ) λ‚΄λ €μ™€μ•Όκ² κ΅¬λ‚˜λΌλŠ” 생각도 λ“€κ²Œ 되죠. νŒ€μ›λ“€λ„ 계속 μ—¬λŸ¬λΆ„μ„ μ‹€λ§μ‹œν‚¬ ν…Œμ§€λ§Œ, κ·ΈλŸΌμ—λ„ 계속 예수처럼 μ‚¬λž‘μ„ νΌμ€˜μ•Ό ν•΄μš”. νŒ€μ›λ“€μ„ μœ„ν•΄ 격리 μ‹œμΌœμ•Ό λ§ˆλ•…ν•œ μ§μ›μ—κ²Œλ„ 웃어야 ν•˜λŠ” μ‹œκΈ°κ°€ μžˆμŠ΅λ‹ˆλ‹€.”
(이승건 비바리퍼블리카 λŒ€ν‘œ, μ •μ£Όμ˜ μ°½μ—…κ²½μ§„λŒ€νšŒ 데λͺ¨λ°μ΄ 2023)
정말 λ¦¬μ–Όν•œ μ°½μ—…μžμ˜ μ‚Ά. μ§€κΈˆ 상황이 λ„ˆλ¬΄ νž˜λ“ λ° 무슨 λΆ€κ·€μ˜ν™”λ₯Ό λˆ„λ¦¬κ² λ‹€κ³  이걸 ν•˜λ‚˜ μ‹Άμ§€λ§Œ 또 λ‚˜λ¦„μ˜ μ΄μœ κ°€ μƒκ²¨μ„œ μ§€μ†ν•˜λŠ” 것이죠. κ°œμΈμ μœΌλ‘œλŠ” 이 삢이 λ‹Ήμ—°ν•˜κΈ°λ³΄λ‹€λŠ” 점차 κ°œμ„ λμœΌλ©΄ μ’‹κ² μŠ΅λ‹ˆλ‹€. μ°½μ—…μžμ— λŒ€ν•œ 쑴쀑과 λ³΄ν˜Έλ„ 컀져야 ν•˜κ³ μš”.
πŸ’―4πŸ€”1
----
Navigating "corporate speak" isn't easy.

Here's a helpful guide I put together:

"Let me check with my team" = No

"Possibly" = No

"On my roadmap" = Not happening

"This will be done in Q4" = This will be done in Q2 next year

"Disagree and commit" = I hate you

"Per my last email" = Try reading, for once in your life

"Challenging landscape" = We're going out of business, quickly

"Digital transformation" = We're going out of business, slowly

"Let's circle back" = We'll never speak of this again

"Take it offline" = We'll never speak of this again

"30,000 foot view" = I don't know what I'm saying

"Low hanging fruit" = Easy promotion

"Open up the kimono" = HR violation

"We use AI" = We don't use AI

"We use machine learning" = We don't use machine learning

"All hands on deck" = Let's actually try for once, please

πŸ˜‚
😁1
아이폰 μΆœμ‹œ ν›„ 첫 4~5λ…„ μ •λ„λŠ” 앱을 μΆœμ‹œν•˜λŠ” κ²ƒλ§ŒμœΌλ‘œλ„ μœ μ €μ˜ wowλ₯Ό μ΄λŒμ–΄ λ‚Ό 수 μžˆμ—ˆλ˜ μ‹œμ ˆμ΄ μžˆμ—ˆλ‹€. μ΄λ™ν•˜λ©° 인터넷도 ν•  수 있고 ν„°μΉ˜μŠ€ν¬λ¦°μ„ ν™œμš©ν•  수 μžˆλŠ” μŠ€λ§ˆνŠΈν°μ—μ„œ ꡬ동이 λœλ‹€λŠ” κ²ƒλ§ŒμœΌλ‘œλ„ wow μ˜€λ˜ 것이닀.

κ·Έ 이후 κΈ°μ‘΄ λ ˆκ°€μ‹œ μ‚°μ—…μ˜ νŒ¨λŸ¬λ‹€μž„μ„ λ°”κΎΌ μ„œλΉ„μŠ€λ“€μ΄ λ“±μž₯ν•˜μ—¬ wow λ₯Ό μ„ μ‚¬ν–ˆλ‹€. 유투브/λ„·ν”Œλ¦­μŠ€λŠ” TV/λ―Έλ””μ–΄ 산업을 λ°”κΏ¨κ³ , 페뢁/ꡬ글은 λ―Έλ””μ–΄/포털산업을 λ°”κΏ”λ†“μ•˜μœΌλ©°, μ•„λ§ˆμ‘΄μ€ μ‡Όν•‘/μ„œλ²„μ‹œμŠ€ν…œμ„, ν…ŒμŠ¬λΌλŠ” μžλ™μ°¨ 산업을, μš°λ²„λŠ” νƒμ‹œ 산업을, 에어비엔비λ₯Ό ν˜Έν…” 산업에 큰 영ν–₯을 쀬닀.

μœ„μ™€ 같은 2000λ…„ 초반 이후 κ²©λ™μ˜ 20λ…„ 정도λ₯Ό κ²ͺ으며, ν˜„λŒ€μ‚¬νšŒμ˜ λ§Žμ€ μ‚¬λžŒλ“€μ€ λͺ¨λ°”일 및 μ›Ή μ€‘μ‹¬μ˜ μ„œλΉ„μŠ€μ— κ½€ 많이 μ΅μˆ™ν•΄μ§„ λ“― ν•˜λ‹€. κ·Έλž˜μ„œ, μƒˆλ‘œ μΆœμ‹œλ˜λŠ” 앱을 봐도 κ·Έ 감ν₯이 κ³Όκ±° λŒ€λΉ„ λ¬΄λŽŒμ§€κ³  μžˆμŒμ„ λŠλ‚€λ‹€.

였히렀, μš”μ¦˜μ€ κ³Όκ±°μ—λŠ” λ§Œμ‘±ν•˜λ©° μ‚¬μš©ν•˜λ˜ Big Tech μ‚¬μ˜ μ„œλΉ„μŠ€ 쑰차도 슬슬 'μ§€κ²¨μ›Œμ§„λ‹€' λŠλΌλŠ” 뢄듀이 μ¦κ°€ν•˜λŠ” λ“―ν•˜λ‹€. λ™μ‹œμ— 'κ³Όκ±°λΆ€ν„° 많이 μ΄μš©ν•˜λ˜ μ„œλΉ„μŠ€λ“€μ΄ μ΅œκ·Όμ—λŠ” μ˜ˆμ „λ§Œ λͺ»ν•΄μ„œ 싀망감이 크고, μ–΄μ©” 수 없이 μ΄μš©ν•˜κ³  μžˆλ‹€'κ³  λ§ν•˜λŠ” 뢄듀도 λ§Žμ•„μ§€λŠ” λ“―ν•˜λ‹€. OpenAI κ°€ chatGPT λ₯Ό μΆœμ‹œν•˜λ©° μƒˆλ‘œμš΄ ν™œλ ₯을 λΆˆμ–΄λ„£κ³  μžˆμ§€λ§Œ, κ·Έ νŒŒκΈ‰λ ₯이 과거의 아이폰 λͺ¨λ©˜νŠΈμ²˜λŸΌ μ—„μ²­λ‚˜μ§€ μ•Šμ€ 것도 사싀이닀. (chatGPTλŠ” μ†ŒλΉ„μž λ³΄λ‹€λŠ” μŠ€νƒ€νŠΈμ—…μ— ν™œλ ₯을 λΆˆμ–΄λ„£κ³  μžˆλŠ” λ“―ν•˜λ‹€. κ·Έλž˜μ„œ 걱정이기도 ν•˜λ‹€. μ†ŒλΉ„μžκ°€ λ°˜μ‘ν•˜μ§€ μ•ŠμœΌλ©΄ κ²°κ΅­ imapct κ°€ μ œν•œμ μΌ μˆ˜λ°–μ— μ—†κΈ° λ•Œλ¬Έμ΄λ‹€) μš”μ¦˜μ€ μœ μ €λΆ„λ“€μ΄ λŒ€λ‹€μˆ˜ μ„œλΉ„μŠ€μ— κ³Όκ±° λŒ€λΉ„ λœ¨λœ¨λ―Έμ§€κ·Όν•œ λ°˜μ‘μ„ λ³΄μ΄λŠ” μ‹œκΈ°μΈ 것 같기도 ν•˜λ‹€.

λ‹€λ§Œ, μœ μ €λ“€μ€ μ—¬μ „νžˆ 더 λ‚˜μ€ 삢을 μ‚΄κ³  μ‹Άμ–΄ν•˜λŠ” μš•κ΅¬κ°€ κ°•ν•˜λ‹€. 더 μ„±μž₯ν•˜κ³  μ‹Άμ–΄ν•˜κ³ , 더 ν–‰λ³΅ν•œ 삢을 μ‚΄κ³  μ‹Άμ–΄ν•˜κ³ , 더 즐거운 μ‹œκ°„μ„ 보내고 μ‹Άμ–΄ ν•œλ‹€. 본인 인생에 κ°•λ ¬ν•˜κ²Œ λ‹€κ°€μ˜¬ 수 μžˆλŠ” μ„œλΉ„μŠ€λ₯Ό λ§Œλ‚˜, ν•΄λ‹Ή μ„œλΉ„μŠ€μ™€ ν•¨κ»˜ 더 만쑱슀러운 삢을 μ‚΄κΈ°λ₯Ό ν¬λ§ν•˜λŠ” 뢄듀이 κ½€ μžˆλ‹€.

κ·Έλž˜μ„œ, λ―Έλž˜μ—λŠ” μ–΄λ–€ μ„œλΉ„μŠ€κ°€ λ§Žμ€ μœ μ €λ‘œλΆ€ν„° wowλ₯Ό μ΄λŒμ–΄ λ‚Ό 수 μžˆμ„μ§€... κΆκΈˆν•˜λ‹€. μœ μ €μ˜ μ„±μž₯κ³Ό λ§Œμ‘±μ— 더 많이 μ§‘μ°©ν•˜λŠ” μ„œλΉ„μŠ€κ°€ κ·Έ wow λ₯Ό μ΄λŒμ–΄ λ‚Ό 수 있으리라 λ―Ώμ§€λ§Œ, κ³Όκ±° λŒ€λΉ„ μœ μ €μ˜ wow λ₯Ό μ΄λŒμ–΄ λ‚΄κΈ° μœ„ν•œ λ‚œμ΄λ„κ°€ λ†’μ•„μ Έμ„œ κ³Όκ±° 만큼 'κΈ‰μ„±μž₯ν•˜λŠ” μŠ€νƒ€νŠΈμ—…'이 많이 λ‚˜μ˜¬ 것 κ°™μ§€λŠ” μ•Šλ‹€. κ·Έλž˜λ„, 이런 λ‚œμ„Έλ₯Ό μ΄κ²¨λ‚΄λŠ” κ°€μž₯ 쒋은 방법은 '더 νž˜λ“€μ–΄μ§€λŠ” ν™˜κ²½'에 λ°˜μ‘ν•˜κΈ° λ³΄λ‹€λŠ” 'μœ μ €'에 μ§‘μ€‘ν•˜μ—¬ ν•˜λ£¨ ν•˜λ£¨ 더 쒋은 μ„œλΉ„μŠ€λ₯Ό λ§Œλ“€μ–΄ λ‚΄λŠ” 것이라 μƒκ°ν•œλ‹€. λ¬Όλ‘  κ³Όκ±°μ—λŠ” 1λ…„ λ…Έλ ₯ν•˜λ©΄ λ§Œμ‘±μ„ μ΄λŒμ–΄ λ‚Ό 수 μžˆμ—ˆλ‹€λ©΄, μ΅œκ·Όμ—λŠ” 3~5λ°°λŠ” 더 ν•΄μ•Όν•˜λŠ” 어렀움이 μ‘΄μž¬ν•˜μ§€λ§Œ 말이닀.

κ·Έλž˜μ„œ μš”μ¦˜ 같은 μ‹œλŒ€λŠ” Why? κ°€ 더 μ€‘μš”ν•΄μ§€λŠ” 것 κ°™λ‹€. μ™œ μ‹œμž‘ν–ˆλŠ”μ§€? 무엇을 μœ„ν•΄ μ„œλΉ„μŠ€λ₯Ό λ§Œλ“€μ–΄ λ‚΄κ³  μžˆλŠ”μ§€?에 λŒ€ν•œ λͺ…ν™•ν•˜κ³  μ†”μ§ν•œ 닡을 κ°€μ§€κ³  μžˆλŠ”μ§€κ°€ 더 μ€‘μš”ν•΄μ§€λŠ” μ‹œλŒ€κ°€ μ™”λ‹€. κ·Έ μ΄μœ μ— λŒ€ν•œ 닡을 전체 νŒ€μ΄ κ³΅μœ ν•˜κ³  μžˆμ–΄μ•Ό, wow λ₯Ό λ§Œλ“€μ–΄ λ‚΄κΈ° μœ„ν•œ λ…Έλ ₯/μ‹œκ°„μ΄ μ¦κ°€ν•œ μ‹œλŒ€λ₯Ό 묡묡히 μ΄κ²¨λ‚˜κ°ˆ 수 있기 λ•Œλ¬Έμ΄λ‹€.

μ„œλΉ„μŠ€λ₯Ό μ΄μ–΄λ‚˜κ°€κ³  μžˆλŠ” 본질적 μ΄μœ λŠ” 무엇인가? 우리 μ„œλΉ„μŠ€λŠ” λˆ„κ΅¬λ₯Ό μœ„ν•΄ μ‘΄μž¬ν•˜λ©° μ™œ μ‘΄μž¬ν•΄μ•Ό ν•˜λŠ”κ°€? κΉŠμ€ λ°€, 링글을 더 μ„±μž₯μ‹œμΌœ λ‚˜κ°€κΈ° μœ„ν•œ 방법을 κ³ λ―Όν•˜λ‹€κ°€, 원둠적인 μ§ˆλ¬Έμ— λŒ€ν•΄ λ‹€μ‹œ ν•œ 번 생각해본닀.
❀3πŸ‘1
β€œMachine learning costs, talent and chip shortages… any AI and machine learning company faces at least one of these challenges, and most face a few at a time,” Pekhimenko told TechCrunch in an email interview. β€œThe highest-end chips are commonly unavailable due to the large demand from enterprises and startups alike. This leads to companies sacrificing on the size of the model they can deploy or results in higher inference latencies for their deployed models.”

With spending on AI-focused chips expected to hit $53 billion this year and more than double in the next four years, according to Gartner, Pekhimenko felt the time was right to launch software that could make models run more efficiently on existing hardware.

β€œTraining AI and machine learning models is increasingly expensive,” Pekhimenko said. β€œWith CentML’s optimization technology, we’re able to reduce expenses up to 80% without compromising speed or accuracy.”

β€œFor one of our customers, we optimized their Llama 2 model to work 3x faster by using Nvidia A10 GPU cards,”

CentML isn’t the first to take a software-based approach to model optimization. It has competitors in MosaicML, which Databricks acquired in June for $1.3 billion, and OctoML, which landed an $85 million cash infusion in November 2021 for its machine learning acceleration platform.

β€œThe CentML platform can run any model,” Pekhimenko said. β€œCentML produces optimized code for a variety of GPUs and reduces the memory needed to deploy models, and, as such, allows teams to deploy on smaller and cheaper GPUs.”
A new solution to the high-end chip shortage.

Read in @WSJ about how Together worked with large former crypto mining farms to repurpose their best GPUs and acquire new GPUs to train AI models β€” all with a specialized training stack for a fraction of the price.

https://www.wsj.com/articles/crypto-miners-seek-a-new-life-in-ai-boom-after-an-implosion-in-mining-92a181fd
μ–Όλ§ˆμ „ ꡬ글 브레인, λ”₯λ§ˆμΈλ“œμ—μ„œ μ½”μ–΄ λ”₯λŸ¬λ‹ νŒ€μ— μžˆμ—ˆλ˜ μΉœκ΅¬μ™€ 이야기 λ‚˜λˆ„λ©΄μ„œ μΈμƒμ‹Άμ—ˆλ˜ λΆ€λΆ„λ“€ (μ΄μ„Έμ’…λ‹˜)

μ§€λ‚œλ²ˆ μ„Έμ…˜μ—μ„œ μΈμƒκΉŠμ€ λΆ€λΆ„μž…λ‹ˆλ‹€.

quantization λ“± λͺ¨λΈ μ΅œμ ν™”λ‚˜ λͺ¨λΈ μ•„ν‚€ν…μ²˜ νš¨μœ¨ν™”λ‘œ λͺ¨λΈ ν•™μŠ΅/인퍼런슀 μ»΄ν“¨νŠΈ(GPU) λΉ„μš©μ΄ μΌμ‹œμ μœΌλ‘œ κ°μ†Œν•  수 μžˆμ§€λ§Œ, μž₯기적 κ΄€μ μ—μ„œ μ»΄ν“¨νŠΈμ˜ μ ˆλŒ€μ  μˆ˜μš”λŠ” λΉ λ₯΄κ²Œ 증가할것
- λ”₯λ§ˆμΈλ“œμ—μ„œ μ•ŒνŒŒν΄λ“œλ₯Ό 3λͺ…μ˜ 연ꡬ원이 ν•΄λ‚Ό 수 μžˆμ—ˆλ˜ 것은 이듀이 더 λ˜‘λ˜‘ν•΄μ„œκ°€ μ•„λ‹ˆλΌ 1인당 ν™œμš© κ°€λŠ₯ν•œ μ»΄ν“¨νŠΈκ°€ λ‹€λ₯Έ κΈ°μ—…, μ—°κ΅¬μ†Œ 보닀 μ••λ„μ μœΌλ‘œ λ§Žμ•˜κΈ° λ•Œλ¬Έ
- μ»΄ν“¨νŠΈκ°€ λ³΄νŽΈν™”λ˜μ–΄ 가격이 μ €λ ΄ν•΄μ§ˆ 수둝 μ•ŒνŒŒν΄λ“œ κΈ‰μ˜ ν˜μ‹ μ΄ λͺ¨λ“  μ‚°μ—…κ³Ό μ˜μ—­μ—μ„œ νŽΌμ³μ§ˆκ²ƒ
- λ‚˜μ•„κ°€ μš°λ¦¬λŠ” μ»΄ν“¨νŠΈμ˜ ν•œκ³„λ‘œ 기본적인 ν…μŠ€νŠΈ 데이터 ν”„λ‘œμ„Έμ‹±μ— 발이 λ¬Άμ—¬ μžˆμ—ˆλŠ”λ° μ•žμœΌλ‘œ μ˜μƒ λ“± λ³΅μž‘λ„ λ†’κ³  무거운 데이터 처리 μˆ˜μš”κ°€ κΈ°ν•˜κΈ‰μˆ˜μ μœΌλ‘œ λŠ˜μ–΄λ‚ κ²ƒ

이미 ν—€μ§€νŽ€λ“œμ—μ„œ μ£Όκ°€ μ˜ˆμΈ‘μ— νŠΉν™”λœ λͺ¨λΈμ„ 적극 ν™œμš©ν•΄ λ§‰λŒ€ν•œ 수읡 μ°½μΆœμ€‘
- λͺ¨λΈ ν•™μŠ΅μ— μ–Όλ§ˆ λ“€κ³ , λͺ¨λΈ κ²½μŸμš°μœ„μ˜ 지속성과 κΈ°λŒ€ λ§€μΆœμ„ κ³ λ €ν–ˆμ„ λ•Œ ROIκ°€ μ–΄λ–»κ²Œ 될지λ₯Ό κ³„μ‚°ν•΄μ„œ λ™μ‹œ μ—¬λŸ¬κ°œ λͺ¨λΈ ν•™μŠ΅μ€‘
- κ·Έμ™Έ λͺ¨λ“  κΈ°μ—…μ—” λ ˆκ±°μ‹œ μ½”λ“œμ™€ 데이터가 μžˆλŠ”λ° λ³΄κ΄€λœ 방식과 μ½”λ“œμ˜ ν˜•νƒœ λ•Œλ¬Έμ— λ§€λ…„ μ–΄λ§ˆν•œ κ³ μ •λΉ„μš©μ΄ λ°œμƒ. AI λͺ¨λΈ 도움을 λ°›μ•„ μ½”λ“œ μ—…λ°μ΄νŠΈμ™€ λ§ˆμ΄κ·Έλ ˆμ΄μ…˜μ„ 톡해 νšŒμ‚¬ μ „λ°˜μ˜ μˆ˜μ΅μ„±μ„ κ°œμ„ ν•˜λŠ” κ²½μš°λ„ 자주 λ³΄μž„

"μ‚¬λžŒμ„ λ•Œλ €λ°•μ•„" μŠ€μΌ€μΌμ„ μΆ”κ΅¬ν•˜λŠ” λΈ”λ¦¬μΈ μŠ€μΌ€μΌλ§ 방법은 AI μ‹œλŒ€μ—μ„œ μœ νš¨ν•˜μ§€ μ•Šμ„ μˆ˜λ„
- κΈ°μ—…μ—μ„œ μ „ν†΅μ μœΌλ‘œ 인건비가 돈이 κ°€μž₯ 많이 λ“€μ–΄κ°”λŠ”λ° μ΅œκ·Όμ— ꡬ글에선 μ»΄ν“¨νŠΈ λΉ„μš©μ΄ 개발자 λΉ„μš©μ„ μ•žμ„œ
- μ˜€ν”ˆAI도 400λͺ… λ˜λŠ” 기업인데 μ»΄ν“¨νŠΈ λΉ„μš©μ΄ 인당 개발자 λΉ„μš©μ˜ 4λ°°
- μ•žμœΌλ‘œλŠ” μž‘μ§€λ§Œ ν”„λ‘œλ•νŠΈλΆ€ν„° μ„ΈμΌμ¦ˆκΉŒμ§€ 전사 μ˜€νΌλ ˆμ΄μ…˜μ— μ–ΌλΌμΈλœ λ‹¨λ‹¨ν•œ νŒ€μ΄ 경쟁λ ₯ μžˆμ§€ μ•Šμ„μ§€

크고 μž‘μ€ μ—¬λŸ¬ μœ ν˜•μ˜ μ œλ„ˆλŸ΄λ¦¬μŠ€νŠΈμ™€ νŠΉν™”λœ SOTA λͺ¨λΈμ΄ κ³΅μ‘΄ν•˜κ²Œλ κ²ƒ
- μœ μ €κ°€ μ›ν•˜λŠ” μš”μ²­μ— λ§žλŠ” λͺ¨λΈμ΄ μžλ™ μΆ”μ²œλ˜μ–΄ νƒœμŠ€ν¬κ°€ μ²˜λ¦¬λ˜λŠ” Model of Experts ν˜•νƒœ
- λͺ¨λΈμ€ 크게 두 μΆ•μ˜ κ΅μ§‘ν•©μœΌλ‘œ μ‘΄μž¬ν• κ±΄λ° ν•œ 좕은 intelligence (예: 80 IQ ~ 150 IQ), 그리고 λ‹€λ₯Έ 좕은 버티컬 (예: μ½”λ”©, 법λ₯ , 의료 λ“±)
- λͺ¨λΈ 검색/μΆ”μ²œ/연결을 λ•λŠ”orchestration λ ˆμ΄μ–΄κ°€ 핡심 기술둜 뢀상할것

좜처: μ΄μ„Έμ’…λ‹˜ 페이슀뢁

https://www.facebook.com/Saejlee
πŸ‘2