Lavic.Kim

Codeshift Inc. CEO/Founder
Entrepreneur, Indi-hacker, Digital Nomad, Trader, Programmer

[Lavic.Kim @LavicKim]

I occasionally trade coin futures. Futures trading has hidden venomous fangs. Sometimes I get bitten, but if I'm careful of these fangs, I can become a skilled tamer. The most important thing is not to get bitten by these poisonous fangs.

[Lavic.Kim @LavicKim]

Technology evokes primal desires

[Jules @julesterpak]

They’re falling in love with Dan (ChatGPT voice)

[Lavic.Kim @LavicKim]

Crazy X, Elon Musk, trying to edit posts and It's pushing for a paid version. Never heard or seen such a crazy payment model before. Such an extreme guy.

(Edit) Devian -> Devin

[Lavic.Kim @LavicKim]

Devian 사용신청 완료.
GPT4가 4명의 일을 해주고 있다만, Devian 너에게는 그 이상을 기대한다.

https://t.co/VGDswUmpRa

[Lavic.Kim @LavicKim]

20 services in 6 month, niche market, AI, primal desires, 16개 남았다.

[Santiago @svpino]

Here's an article that provides a step-by-step guide for fine-tuning the LLMs and more details about their platform:

https://t.co/pcH33TWB3I

Join their Discord server for the latest updates, free credits, and special offers:

https://t.co/oecoxEec14

Thanks to the team @monsterapis for partnering with me on this post.

[Santiago @svpino]

GPT-4 is getting worse over time, not better.

Many people have reported noticing a significant degradation in the quality of the model responses, but so far, it was all anecdotal.

But now we know.

At least one study shows how the June version of GPT-4 is objectively worse than the version released in March on a few tasks.

The team evaluated the models using a dataset of 500 problems where the models had to figure out whether a given integer was prime. In March, GPT-4 answered correctly 488 of these questions. In June, it only got 12 correct answers.

From 97.6% success rate down to 2.4%!

But it gets worse!

The team used Chain-of-Thought to help the model reason:

"Is 17077 a prime number? Think step by step."

Chain-of-Thought is a popular technique that significantly improves answers. Unfortunately, the latest version of GPT-4 did not generate intermediate steps and instead answered incorrectly with a simple "No."

Code generation has also gotten worse.

The team built a dataset with 50 easy problems from LeetCode and measured how many GPT-4 answers ran without any changes.

The March version succeeded in 52% of the problems, but this dropped to a pale 10% using the model from June.

Why is this happening?

We assume that OpenAI pushes changes continuously, but we don't know how the process works and how they evaluate whether the models are improving or regressing.

Rumors suggest they are using several smaller and specialized GPT-4 models that act similarly to a large model but are less expensive to run. When a user asks a question, the system decides which model to send the query to.

Cheaper and faster, but could this new approach be the problem behind the degradation in quality?

In my opinion, this is a red flag for anyone building applications that rely on GPT-4. Having the behavior of an LLM change over time is not acceptable.

Have you noticed any issues when using GPT-4 and ChatGPT lately? Do you think these problems are overblown?

[anton @abacaj]

We are so early

[LangChain @LangChain]

✒️Blog

what we built, where we’re going, and how our Alpha partners put LangSmith to use

🙏🏽 to companies like @klarna @SnowflakeDB, @streamlit @BCG @DeepLearningAI_ @fintual @mendableai @multion_ai & @quivr_brain for helping us shape LangSmith

https://t.co/s4QuaHyMHb

[CalCo @calco_io]

Current AI technology is concentrated in the hands of a very small number of companies with outsized access to compute resources. We are changing that:

Powerful models should be accessible to all: https://t.co/28GSI4XBcF
The tools to build them as well: https://t.co/nChw9h8pmn

[Replicate @replicate]

Language models are now on Replicate.

Open-source language models have been improving at an extraordinary pace. You can now deploy, run, and fine-tune them in the cloud with just a few lines of code.

https://t.co/iD1Mjxlz1G

[Micah's Gambit♟️ @Liu3King]

The picture is alive! DragGAN has an open source Star volume of 23k in three days.
#AIGC #LLMs
https://t.co/BNB42o8gk0

[Lavic.Kim @LavicKim]

"B2B SaaS + LLM = Nimble Powerhouse. Such a great world for startups!"

[Lavic.Kim @LavicKim]

B2B SaaS + LLM = Nimble Powerhouse

https://t.co/oL0fhzcWMt

[Lavic.Kim @LavicKim]

[바라던 바다▶모아듣기] 석양도 로제(ROSÉ) 앞에선 조명일 뿐,, 월클 로제 모아듣기(무대.ver) | JTBC 210803 방송 https://t.co/WOS8ZZshFb 출처 @YouTube

[Lavic.Kim @LavicKim]

🎤 밤바다를 감성 가득 하모니로 채운 하림(Hareem)x소향(Sohyang) ′I′ll Be There′♪ 〈비긴어게인 코리아(... https://t.co/fCuY6sad4G 출처 @YouTube

[Lavic.Kim @LavicKim]

Cupid 💘 Song #shortsvideo #fypシ゚viral #trendingshorts #songs #cupid #fif... https://t.co/hGXBqKzlpL 출처 @YouTube

[Lavic.Kim @LavicKim]

[아란] 빌보드 그 노래 피프티피프티 큐피드 라이브 #shorts https://t.co/Qh6wOHmZm4 출처 @YouTube

[Lavic.Kim @LavicKim]

1 그룹 4 보컬 화음 미쳤다… https://t.co/ot6cgn8N2Q 출처 @YouTube

[Lavic.Kim @LavicKim]

[Sohyang/소향] 어사널사 부르는 소향의 놀라운 음색 ( Sohyang's Amazing voice) #Shorts https://t.co/tkF0TfeTRg 출처 @YouTube