Continuous Learning_Startup & Investment – Telegram
Continuous Learning_Startup & Investment
2.4K subscribers
513 photos
5 videos
16 files
2.72K 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
🔍Reddit API 가격 논란 심층 분석(Feat. ChatGPT & Perplexity)
빠르게 변화하는 디지털 세상에서 최신 소식을 파악하고 이해하는 것은 종종 벅찬 일입니다. 최근 논란이 되고 있는 Reddit의 API 가격 변경 결정이 그 예입니다. 오늘은 Chat GPT와 같은 AI 도구를 사용하여 뉴스를 분석하고 주요 인사이트를 수집하여 이 사건을 자세히 살펴보고자 합니다.

먼저 문제 자체부터 살펴보겠습니다. 최신 사실을 확인하기 위해 실시간 정보 검증에 탁월한 플랫폼인 Perplexity를 활용했습니다. 자세한 링크는 이 링크를 확인해주세요. https://www.perplexity.ai/.../22f0415f-5df5-454c-a6b3....

2023년 7월 1일부터 Reddit은 타사 애플리케이션에 API 호출에 대해 1,000회 호출당 0.24달러의 요금을 부과하기 시작할 예정입니다. 사소한 금액처럼 들릴 수 있지만, Apple 기기 소유자가 널리 사용하는 Apollo와 같이 트래픽이 많은 앱을 고려하면 그 금액은 천정부지로 치솟습니다. 한 달에 약 70억 건의 요청이 발생하면 연간 2천만 달러라는 엄청난 비용이 발생할 수 있습니다. 당연히 개발자와 사용자 모두 비판적인 반응을 보였고, 여러 하위 레딧에서 항의의 물결이 일었으며, 일부 레딧은 이에 대한 대응으로 사이트가 폐쇄되기도 했습니다.

이에 대해 Reddit의 CEO인 스티브 허프먼은 Reddit이 자립형 비즈니스로 발전해야 할 필요성을 지적하며 이번 결정을 정당화했습니다. 그러나 많은 사람들은 이러한 허프만의 입장을 타사 앱과의 경쟁에서 우위를 점하기 위한 노력으로 보고 있으며, 트위터의 유사한 조치와 유사합니다. 이에 대한 반발은 2023년 6월 12일 3시간 동안의 Reddit 서비스 중단으로 이어졌습니다.

Perplexity가 현황에 대해서 자세히 설명해줬다면, 이제 Chat GPT 4를 활용하여 이 상황에서 중요한 인사이트를 추려보고 Reddit CEO가 사태를 수습하기 위해서 무엇을 해야하는지 이야기해보겠습니다.

제가 사용한 프롬프트는 다음과 같습니다:
Role: Serial entrepreneur who have dealt with controversial and sensitive issues for a company.
Task: Derive major insight we can learn from <Reddit's problem> and <response from the twitter community>. And suggest what should have Reddit CEO did? And if you were Reddit CEO right now what could he consider mitigating the problem?

1. 이 사례를 통해서 무엇을 배울 수 있나요?
1. Reddit 사건과 트위터 커뮤니티의 반응을 통해 몇 가지 주요 인사이트를 도출할 수 있습니다:
2. 비즈니스 의사 결정 대 커뮤니티 정서: Reddit의 API 가격 변경에 대한 반응은 서비스를 통해 수익을 창출하려는 기업의 필요성과 접근성을 유지하려는 사용자 커뮤니티의 욕구 사이의 갈등을 보여줍니다. 많은 사람들이 이러한 조치가 타사 앱을 단계적으로 폐지하려는 의도적인 시도로 간주하여 반발이 일어났습니다.
3. 투명성과 소통: Reddit의 CEO인 스티브 허프먼의 내부 메모와 공개 성명은 상황을 과소평가한 것으로 간주되었습니다. 그의 메시지는 개발자와 사용자들이 제기한 우려를 해결하지 못했습니다. 이는 위기 상황을 관리할 때 명확하고 정직한 커뮤니케이션의 중요성을 강조합니다.
4. 브랜드 이미지와 신뢰에 미치는 영향: 주요 서드파티 앱인 아폴로와 RIF에 개입하지 않기로 한 결정은 사용자와 개발자 간의 신뢰를 더욱 약화시켰습니다. 그 결과 Reddit의 브랜드 이미지에 부정적인 영향을 미쳤습니다.


2. Reddit CEO는 어떤 의사결정을 다르게 내렸어야 이 문제가 없었을까요?
1. 이해관계자를 조기에 참여시키세요: 이러한 중대한 변경을 단행하기 전에 허프먼은 개발자 커뮤니티 및 기타 주요 이해관계자에게 연락하여 제안된 변경 사항, 잠재적 영향, 가능한 완화 전략에 대해 논의했어야 했습니다.
2. 보다 명확하고 공감할 수 있는 커뮤니케이션: 내부 메모는 논란을 정면으로 다루고 영향을 받는 사람들에 대해 더 많은 공감을 표시했어야 했습니다. 단순히 분노가 지나갈 것이라고 말하는 대신, 허프먼은 불만을 인정하고 우려 사항을 해결하기 위한 구체적인 단계를 자세히 설명했어야 했습니다.
3. 점진적인 도입 또는 대안적인 가격 구조를 고려하세요: 허프만은 갑작스럽게 높은 비용을 도입하는 대신 API 가격을 점진적으로 인상하거나 사용량에 따라 계층화된 가격 모델을 도입하는 방안을 고려할 수 있었을 것입니다. 이렇게 하면 개발자의 즉각적인 재정적 부담을 줄일 수 있었을 것입니다.


3. 지금 Reddit의 CEO로서 문제를 완화하기 위해 무엇을 해야하나요?
1. 가격 모델을 재평가합니다: Reddit의 수익 필요성과 개발자의 서비스 이용 능력 간의 균형을 맞출 수 있는 보다 유연한 가격 구조를 모색하세요.
2. 커뮤니티와의 열린 대화: Reddit 커뮤니티와 진정성 있는 대화를 통해 커뮤니티의 우려 사항, 제안 사항, 기대치를 파악하세요. 투명성은 신뢰 회복에 도움이 됩니다.
3. 개발자를 지원하세요: 가장 큰 영향을 받은 개발자에게 재정적 지원 또는 전환 지원을 제공하는 것을 고려하세요. 이는 선의와 Reddit이 개발자의 기여를 소중히 여긴다는 것을 보여줍니다.
4. 아폴로와 RIF에 대한 결정을 재검토합니다: 아폴로와 RIF를 구하기 위해 개입하지 않기로 한 결정을 재고합니다. 해당 앱의 서비스 종료 가능성에 대한 부정적인 정서가 상당하며, 이를 해결하면 대중의 인식을 개선하는 데 도움이 될 수 있습니다.
<투자 한글번역 영상 셀렉션>

1. 찰리 멍거 | 인생을 살아가는 데 도움이 되는 몇 가지 조언들
- https://youtu.be/efgIt8MxZ4c

2. 세콰이어 캐피탈 회장 마이클 모리츠, 변화와 흐름에 대한 통찰
- https://youtu.be/OwkaChNCVWE

3. 애스워드 다모다란 : 회사를 밸류에이션 하는 방법
- https://youtu.be/PI5YyNFPDZA

4. 비노드 코슬라 인터뷰 | 썬마이크로시스템즈 창업자, 코슬라벤처스 창업자
- https://youtu.be/Hzfrdc221nc

5. 나심 탈레브 : 그 누구의 조언도 듣지 말고, 내면의 목소리를 듣고 따라야 한다
- https://youtu.be/DFyRT_M5f4k

6. 주식시장을 이기는 투자 | 피터린치 | Peter Lynch
- https://youtu.be/ogNL5buu8JA

7. 워렌버핏이 본질을 바라보는 법
- https://youtu.be/l39su7Ty9ek

8. 피터틸 - "기술의 저주" | The Tech Curse
- https://youtu.be/wzkZi6ROFy0

9. 워렌버핏과 찰리멍거 | 2023년에 남기는 삶의 조언들
- https://youtu.be/gSSzeMPpU3c

10. 구글, 페이스북, 트위터, 페이팔 전부 '사람만 보고 투자했습니다'
- https://youtu.be/wDPZuH3Q_qw
Forwarded from BZCF | 비즈까페
BZCF 텔레그램 : https://news.1rj.ru/str/bzcftel

유니클로 야나이 다다시 회장의 2005년 직원들에게 작성한 이메일의 일부입니다. 제가 메모처럼 적어두고, 주기적으로 읽어 보는 글입니다.

2005년은 다다시 회장이 다시 유니클로로 돌아왔던 시기였습니다. 당시 유니클로는 적당히 큰 대기업이 되었고, 사람들은 '적당히 하자' 주의가 팽배했었습니다. 규모도 ‘적당히’ 컸다는 것이, 현재의 유니클로와 같이 글로벌한 회사는 아니었고 정말로 '적당히'였습니다.

그런 시기에, 제2의 창업을 선언하고, 기존의 것들을 완전 새로 하자는 말을 했었고요. 그 후 20년이 지나서... 패션 업계에서 유일무이한 아시아 최초로 글로벌한 회사가 되었네요. 없던 것을 창업하는 것도 어렵지만... 이미 세팅된 문화를 갈아 끼우는 것이 더 힘들었을 것이라 생각이 듭니다. 그의 책을 다시 펼쳐 읽다가, 이 구절이 인상 깊게 다가와 필사하였습니다.

https://m.blog.naver.com/bizucafe/223128943351
회사의 크기는 리더의 꿈의 크기로 수렴한다.
2
theinformation.com/articles/why-youtube-could-give-google-an-edge-in-ai

Summarize this contents and derive key insights as a ai related startup founder and investor via Perplexity copilot feature.

#1 As an AI startup founder and investor, here are some key insights from the provided search results:
Google's ownership of YouTube could give it an edge in AI. YouTube is the single biggest and richest source of imagery, audio, and text trannoscripts on the internet, and Google's researchers have been using it to develop its next large-language model, Gemini. OpenAI has also secretly used data from YouTube to train some of its AI models.

#2 Open-source AI is gaining on Google and ChatGPT. Meta Platforms gave academics access to sophisticated machine-learning models that can understand conversational language, which they turned into open-source software that powered free alternatives to ChatGPT and other proprietary AI software. Free AI models are now "reasonably close" in performance to proprietary models from Google and ChatGPT creator OpenAI, and most software developers will eventually opt to use the free ones

#3 Google and DeepMind are working together to develop software to compete with OpenAI. Known internally as Gemini, the joint effort began in recent weeks, after Google stumbled with Bard, its first attempt to compete with OpenAI's chatbot

#4 Generative AI startups need their own data. Google, Amazon, Meta Platforms, and Microsoft are jostling to develop products that use generative AI. Startups such as Denoscript, which uses generative artificial intelligence to shortcut the laborious video-editing process, are trying to immunize themselves from looming big tech competition

#5 The European antitrust case against Google might or might not matter. The Europeans have concluded, in a preliminary way, that Google favors its own "online display advertising technology services to the detriment of competing providers". This is no small matter, as Google controls the tools most commonly used by both websites selling ad space and marketers buying ad space on websites across the internet. However, it remains to be seen whether this case will have a significant impact on Google's AI-related activities.

#6 Google is preparing to teach foreign languages through Google Search. The project, referred to internally as Tivoli, grew out of its Google Research unit and is likely to be rolled out later this year.

#7 The creator economy is facing challenges and opportunities with AI. Dmitry Shapiro and Sean Thielen, the co-founders of a startup that developed Koji, a tool that offers link-in-bio landing pages to house creators' digital storefronts, are planning to focus on a new AI product they're developing called YouAI. However, venture capitalists such as Ben Mathews are predicting a wave of failures among creator economy startups as they run out of cash
Continuous Learning_Startup & Investment
https://youtu.be/rYVPDQfRcL0
AMD has revealed the Mi 300X chip, which has an industry-leading 192 GB memory capacity, 5.2 TB per second memory bandwidth, and is designed for generative AI. It reduces the number of GPUs required and development time needed for deploying the Mi 300X while accelerating customers' time to market, reducing overall development costs and making deployment effortless. The Mi 300A is currently being sampled while the Mi 300X and eight GPU Instinct platform will begin sampling in Q3, with production expected in Q4 of this year.
Continuous Learning_Startup & Investment
https://youtu.be/ltQ9pbFukUo
Chris Lattner and Lex Fridman discuss the potential of large language models (LLMS) in programming, including their ability to predict and generate code. While LLMS can automate mechanical aspects of coding, it is not a replacement for programmers but a helpful complementary tool. The discussion also covers the potential for LLMS to improve productivity and learn different programming languages. Lattner notes that LLMS can be used for documentation and inspiration, but creating reliable scale systems should focus on algebraic reasoning and creating different nets to implement code rather than expensive LLMS.
Whenever I use an AI product that offers a great user experience, I ask myself: "How the F#🤬 did they do it?"

And if the product is open source, I dive deep into the code, the logic, and of course, the almighty prompt.

Today, I was exploring Bamboo AI, a tool that analyzes data using natural language processing.

Being the curious geek that I am, I dove deep into the prompt to see how it was engineered and to see what types of techniques I could pick up to make my own prompts better.

And in this tweet, I will do my best to reverse engineer the prompt into its building blocks.

Feel free to bookmark this tweet for later reference. I've broken down the prompt into simple pieces for you to replicate if you want.

PS: A (slighlty reduced) snippet of the prompt is attached on the images for reference.
Here's what I found:

1. Role

The prompt starts by declaring a clearly defined role for the AI. Most prompts do this, as this has become a standard best practice.

2. Goal

A clearly defined goal on top of the role allows the AI to act in accordance to it. Nothing extraordinary with this prompt until now.

The problem is that most people stop crafting their prompts here, and then wonder why their outputs are useless and random more often than not.


3. Clearly defined input

The prompt clearly states what is the expected input the AI will receive.

This part of the prompt is often overlooked, but I've found it greatly reduces the randomness of the output.

4. Clearly Defined output

Similarly, clearly going over the expected output in minute detail will help steer the model in the exact direction that you want.

This will allow you to pinpoint exactly what it should do, and on top of that, will reduce the need for revisions.

Again, most people never even get to this point and then wonder why the AI never gets them right.

Of course it doesn't; it won't get you right if you haven't told it what to do.

5. Revisions

The prompt clearly states that revisions are to be expected and the output probably won't be the final one.

Once again, clearly stating what can happen during the actual use of this tool.

6. Input example

On top of clearly defining what input to expect, the prompt also shows an input example.

"Show, don't just tell" is a good principle to keep in mind when prompting.

This will greatly reduce the randomness of the model and make for more accurate outputs.

7. Output example

Stating the examples of the output is equally important.

This will allow the model to pick up on the input -> output pattern and make its answers way more relevant, contextual and useful.

See a pattern here?

Clearly state what the AI should do and what to expect.

Don't leave it to chance if you want your outputs to be reliable and useful.

And now that we have reverse engineered how this prompt works, you will hopefully have ideas on how to improve your own prompts.

I sure did.

https://twitter.com/Luc_AI_Insights/status/1668792631806050304?s=20