Another day, another major OpenAI model announcement. This one is big. Today's GPT-4o announcement (the "o" is for omni) introduced a single AI model that can reason across audio, vision, and text at once. What's incredible about this model is that it has made working through different information and media types incredibly seamless and highly performant, including some amazing new consumer product experiences in ChatGPT. But for anyone building software in AI, one of the biggest components was the drop of GPT-4o pricing by 50% and the doubling of performance compared to GPT-4 Turbo.

That means that any developer building on GPT-4 woke up today and essentially had the cost of intelligence drop by 50% and access to it speed up by 2X. And incredibly, across numerous benchmarks, GPT-4o appears to be the best performing model generally available right now. The implications of this are massive, as this opens up yet another new threshold of use-cases that you can be building for due to the cost/performance improvement. It also is yet another reminder that you want to have a platform architecture that ensures you're getting access to all the latest AI breakthroughs no matter where they come from.

https://plus.hankyung.com/apps/newsinside.view?aid=202310045582r&category=home&sns=y

테크 분야 혁신 기업들은 운영 차원에서 조금만 도움을 받으면 큰 현금 흐름을 낼 수 있는 회사들입니다. 저희가 23년째 해온 일은 혁신을 가지고 있는 회사들을 수익을 낼 수 있는 회사로 변모시켜주는 것입니다. 그리고 그 수익을 가지고 다시 더 빠른 성장이라든지 연구개발이라든지 엔지니어링 연구라든지 하는 곳에 재투자를 할 수 있도록 도와주는 파트너십을 맺는 것이 저희의 노하우입니다. 지금까지 이런 간단한 공식을 지금까지 계속 반복해서 적용하고 있습니다.

저희는 하나의 어젠다를 세우고 기존 창업자와 경영진 그리고 임직원과 함께 일하는 것을 선호합니다. 경험이 누적된 파트너들이 비즈니스 요소를 분석하면 저희가 각 요소에 적합한 투자를 할 수 있기 때문입니다.

As someone who spent a lot of time making a browser and researching it, I can tell you that this integration of ChatGPT on to the computer belies a greater purpose—one where AI will eat the browser steadily. They will no longer have to be restricted by the Google's platform limits.

It’s not just the browser , AI is going to kill “browsing the web” as a paradigm. We are no longer in “search browse and find” era.

We are entering into an era of “ask and get”.

Jack’s first piece of advice is to make sure you have something to show investors:

“First and foremost, I always want to go to anyone that I want to work with with something to show.”

At Twitter, they made sure they had people using the service at some scale before pitching investors. And at Square, they had 7 merchants and a working prototype that people were actually using.

As Jack explains:

“If people can’t see it and feel it, it’s very hard to sell. And I don’t want to just go and tell people this is going to be the biggest thing ever. I want to be able to show that and have them feel it and walk away feeling that as well.”

Jack’s second piece of advice is to make sure you really like the person you’re raising money from because ultimately investors are like employees you can’t fire.

Ask yourself:

“Do we really want to work with this person? Is this going to to be someone that really pushes us in the way we need to be pushed?”

Jack recounts pitching Square early on and getting several term sheets from VCs that asked zero questions:

“I said, I don’t want a term sheet—especially if you’re not asking any questions. That means if they’re on our board that they’re actually not going to be all that constructive in terms of really asking the tough questions. That is the role of the board. It’s to look outside and bring all their insight and wisdom to us and help guide the company.”

With Square, Jack decided to work with Vinod Khosla who asked tough questions and after he invested would email Jack once a week asking things like: “Have you thought about this?”, “Are you doing this?”, “I know this person, can I make an introduction here?”

When people ask Jack which firms they should go pitch, Jack responds:

“It’s not just about money or the firm. It’s about the individual that you’re ultimately going to work with… Find the people you really want to work with in those firms. And if there’s someone that just really resonates with you and you love the idea of working with them, focus on them. You have to treat it as adding this person to your team.”

갑자기 유명해지는 사람은 경계할 필요가 있다는걸 다시 한번 상기해본다.

물론 글에서도 나와있는 것 처럼 차마스는 엄청난 성과를 거둔 사람인건 맞으나, 그 성과가 지속가능한 것일지는 다른 문제인 것이다.

계속해서 잘하는건 정말 정말 어려운 일이다. 그런 의미에서 최근에 버크셔 해서웨이 재탄생 + 버핏 투자조합 서한을 다 읽어보니 버핏은 레전드 of 레전드라는 생각밖에 들지 않았다.

https://www.newcomer.co/p/the-dictator-chamath-palihapitiyas

1/ Weekly 1:1s with direct reports are a staple of Silicon Valley management. The idea is to check in, see how they're doing, and provide feedback. I did this for 10+ years at Facebook & Dropbox. Frankly, I hated it and found it useless. But it's what "good" managers did.

2/ I'm now convinced these regular 1:1s do more harm than good. The simple reason: they condition people to do spot checks on happiness and constantly be critical about things that aren't ideal. In practice, 1:1s descend into nitpicking sessions.

3/ I want my employees to be resilient. To understand that not every week or month will be happy and pleasant. I want them to deal with it without constantly feeling bad. Weekly 1:1s undermine this.

4/ This doesn't mean feedback isn't useful. But it should be given every 3-6 months, not weekly. This forces managers to identify patterns and provide holistic guidance, rather than doing weekly "spot checks."

5/ Using 1:1s to track what someone is working on is also archaic. There are plenty of modern methods to assess output across a range of functions. Wasting time on status updates is inefficient.

6/ A better model is to be generally available over email/Slack for questions. Then have deeper career development conversations over a meal once a quarter or so. This is a more effective cadence.

7/ Save yourself and your reports time and emotional energy to focus on the things that really matter - executing and making the company successful. Excessive 1:1s are a distraction.

8/ If an employee is really struggling, then of course have more frequent check-ins. But for most folks, quarterly big picture conversations and real-time availability are sufficient.

9/ I know I'm arguing against orthodoxy here. But I believe the relentless cadence of weekly 1:1s, while well-intentioned, has become counterproductive. It's time to break free of this flawed Silicon Valley management "best practice." Your team will thank you.

https://x.com/adityaag/status/1790798956886438278?s=46&t=h5Byg6Wosg8MJb4pbPSDow

오픈AI에서 이제 Google Drive 또는 Microsoft OneDrive에서 직접 다양한 파일 형식(구글시트, 엑셀, 워드 등)을 추가할 수 있도록 했습니다.

또한 전체화면으로 GPT와 대화를 통해 데이터 분석을 할 수 있다고 합니다😀

https://openai.com/index/improvements-to-data-analysis-in-chatgpt/

이번 실적발표 시즌의 주인공인 실리콘투와 삼양식품. (뻔하지만) 결국 국내 회사들의 답은 글로벌이라는걸 다시 한번 확인할 수 있었다.

(개별종목 이야기 거의 안하는데, 이번껀 의미가 좀 남다른 것 같아서 아카이브...)

https://blog.naver.com/tosoha1/223449281696

https://blog.naver.com/foreconomy/223449103042

법률 + AI 서비스가 또 등장했네요. 시드 라운드에 세상에서 투자 최고 잘한다고 생각하는 곳들이 붙었네요.

저는 AI가 붙는데 있어서 법률 산업이 코딩 다음으로 활발한 분야라고 여전히 생각합니다.

"Leya, a Stockholm, Sweden-based AI-powered legal workflow platform, raised $10.5 million in seed funding. Benchmark led the round and was joined by Hummingbird, SV Angel, and Y-Combinator."

https://www.leya.law/

칩이 이식된 후 놀라운 일이 벌어졌다. 초기에는 생각을 행동으로 하는 방법을 배워야 했다. 하지만 얼마 지나지 않아 체스를 두면서 사람들과 대화하는 것까지 가능해졌다. 마치 슈퍼히어로처럼 초능력을 가진 것 같았다. 휠체어에 앉아 펜을 입에 물고 아이패드를 사용할 필요 없이 생각하는 대로 컴퓨터의 여러 창을 옮겨 다니고, 시드마이어의 문명이나 체스를 마치 손으로 직접 하는 것처럼 플레이할 수 있었다.
하지만 최근 문제가 발생하기도 했다. 그의 생각과 컴퓨터 사이에 지연이 생기고 성능이 떨어지기 시작했다. 뉴럴링크는 뇌에 전극을 연결한 실이 느슨해졌고 이것이 성능 저하로 이어졌다고 밝혔다. 다만 정확한 원인이 무엇인지는 공개하지 않았다고 블룸버그는 보도했다.
아르보는 “한 번 뉴럴링크를 사용해보면 멈출 수 없다”면서 “1년 간의 계약이 끝난 후에도 최신 모델을 사용하고 싶다”고 말했다. 놀런 아르보가 계속 칩을 사용하게 될지는 1년 후에 결정된다.
아르보는 “내가 생각하는 말을 그대로 컴퓨터에 입력해 줄 수 있었으면 좋겠다. 판타지 소설 작가가 되는 것이 꿈이다”라고 블룸버그에 말했다. 그는 최초의 뉴럴링크 이식 환자가 된 것이 그가 가족에 덜 의존하는 기회가 되길 원한다고 설명했다.
아르보는 “내 남동생은 나를 8년이나 돌봤다. 그도 자신의 인생을 살아야 한다. 그동안 돌봐준 어머니를 위해서 집을 지을 수 있을 만큼의 돈을 벌고 싶다”고 말했다.

https://www.mk.co.kr/news/society/11017864