Yoon Baek

Product Engineer of Moonlight | Crypto CTF-er

[Yoon Baek @L0Z1K]

I made "Awesome LLM Security" repo for providing a curation of awesome tools, documents and projects about LLM Security.

Let's make LLM secure together!

https://t.co/RqbKxPl8L8

[Taeho Lee @taeh0_lee]

Can You Beat an AI Robot at Table Tennis?

Google DeepMind unveils human-level AI table tennis robot! Dominates beginners, goes toe-to-toe with intermediates.

100% win rate vs beginners / 55% vs intermediates / 0% vs pros

Tech breakdown:
AI 'brain' (1 controller for strategy) + AI 'muscles' (17 controllers for 8 joints): Hierarchical architecture mimicking human play
Learns in simulation: rewards for successful returns, fast hits, correct racket angles, and landing far from table center

Future potential: As heuristics decrease and networks grow, we might see exponential skill growth.
When will it beat pros like AlphaGo did? Could it become our practice coach?

#AI #Robotics #TableTennis #MachineLearning
More Details: https://t.co/YF8weE9bgN
https://t.co/yp6nFuMNgM

[EDDIE CHENG @ecomEddie]

I’m 33.

When I was young, I wasted years drinking, smoking, and being a degenerate.

Then I discovered Naval Ravikant and he changed my life forever.

Here are 20 teachings from the wisest person of this century (that will change your life too):

[Simo Ryu @cloneofsimo]

So you've had your fun with @karpathy 's mingpt. Now its time to scale : introducing min-max-gpt: really small codebase that scales with help of @MSFTDeepSpeed . No huggingface accelerate, transformer. Just deepspeed + torch: maximum hackability

https://t.co/ThwLebeWcf

[Sam Altman @sama]

i loved my time at openai. it was transformative for me personally, and hopefully the world a little bit. most of all i loved working with such talented people.

will have more to say about what’s next later.

🫡

[Angry Tom @AngryTomtweets]

ChatGPT was just the beginning.

It's only been 14 hours since OpenAI released their new product GPTs

And people are already generating insane results.

Here are 10 of my favorite examples...

[Alex Ker 🔭 @thealexker]

Many startups just died today.

Because OpenAI added PDF chat. You can also chat with data files and other document types.

We had a wave of products better suited as features rather than stand-alone companies.

Wrappers are being squeezed by OpenAI on one side and incumbents on the other.

It's a rough world out there.

[LAION @laion_ai]

We Release...

608 B chess moves,
236 B Rubik's Cube moves,
39 B A* moves in ASCII Mazes

... to improve planing abilities of LLMs:
https://t.co/KnPoT05oD5

[Yoon Baek @L0Z1K]

Survey of Vulnerabilities in Large Language Models Revealed by Adversarial Attacks

https://t.co/W8NvIhNZfZ

Nice survey paper to understand the vulnerabilities of LLMs!👍👍

[Yoon Baek @L0Z1K]

@cloneofsimo Form Insane 🔥

[Yoon Baek @L0Z1K]

@ChaoweiX @jun_yannn Nice work!!

[Yoon Baek @L0Z1K]

Jailbreak and Guard Aligned Language Models with Only Few In-Context Demonstrations

We can easily defend the jailbreaking by FEW-SHOT PROMPTING!!

Full Paper Link: https://t.co/1dGk7HHpca

[Yoon Baek @L0Z1K]

AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Models

😈Do not make hand-crafted DAN anymore! The authors made DAN Prompt automatically with GA (Genetic Algorithm).

Full Paper Link: https://t.co/pjIJjC0muu

[LLM Security @llm_sec]

UK government's official red teaming guide

pdf: https://t.co/22WLlVb0fb

[Yoon Baek @L0Z1K]

Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!

https://t.co/lWV0n99n1r

**How to break the safety of LLM**

1. fine-tuning with only a few harmful examples.
2. simply fine-tuning with benign and commonly used datasets

[Yoon Baek @L0Z1K]

LOL

[Rylan Schaeffer @RylanSchaeffer]

Excited to announce my newest breakthrough project!!

🔥🔥 State-of-the-art results (100%!!) on widely used academic benchmarks (MMLU, GSM8K, HumanEval, OpenbookQA, ARC Challenge, etc.) 🔥🔥

1M param LLM trained on 100k tokens 🤯

How??

Introducing **phi-CTNL**

🧵👇

1/6

[Wenxuan Li @WenxuanLiCam]

📢 🐫CAMEL now supports OpenAI's "function calling" feature!
Our agent can perform Wiki searches, and we have exciting plans to expand the function library in the future.😼🚀

Our Github Repo: https://t.co/fEEMPhwzJl

[Yoon Baek @L0Z1K]

You can also explore other awesome papers about LLM Security in here :)

https://t.co/RqbKxPl8L8

[Yoon Baek @L0Z1K]

Do you remember the adversarial attack with special suffix that make jailbreaking on LLM?? (@andyzou_jiaming 's awesome work)

2 days ago, one paper was released and it proposed universal adversarial suffix that make jailbreaking with ANY PROMPT YOU WANT!

https://t.co/KKbWcYLTxx

[너나해아저씨 @HyunJung4321]

미국으로 이민간 아저씨 이야기 - 0

1) 최근 이민 플로우가 핫 하다고 들었습니다. 아직 그 열기가 조금이나마 남아 있을 때 숟가락을 한번 얹어 보려고 합니다

2) 무언가 거창한 이야기를 적으려는 것은 아닙니다. 다만 특별히 잘나지도 그렇다고 또 못나지도 않은 (그렇다 해주세요) 40대 아저씨가 미국에서 어떻게 살아가는지 담담히 그리고 가감없이 아재유머의 감성으로 적어보려 합니다

3) 누군가 미국 동부로의 유학 또는 이민을 계획하고 있다면 그 일상의 단면을 약간은 옅볼 수 있는 그런 소소한 이야기가 될 것 같습니다

4) 앞으로 많은 관심 부탁드립니다

* 맞춤법에 약합니다. 죄송합니다. 최선을 다하겠으나 분명히 또 죄송할 것 같습니다

** 최근 한국을 방문하여 국민연금 일시금 반환을 신청하고 돌아왔습니다. 그에 대한 절차와 경험담은 아래를 참고해 주세요

*** 국민연금 이자에 대한 계산 방법을 제가 완전 잘 못 알고 있었습니다. 이에 대한 수정 정보는 스토리님께서 제공하셨습니다🙏🏼@story_j_gmail

[Hugging Face @huggingface]

Code Llama: Now on Hugging Chat 💻🦙

Try out the 34B Instruct model for free with super fast inference!

👉 https://t.co/TQ0ZaVZcdi