abdel

@AbdelStark

Director of Research Partnerships & Standards at @LucidCompute. AI Verification | AI Safety | ZK

🤗 huggingface.co/abdelstark

github.com/AbdelStark

2012年3月からXを利用しています

6,546 フォロー中

3万フォロワー

固定
abdel
@AbdelStark
11月4日
Some personal news: I joined Lucid Computing (@LucidCompute) as Director of Research Partnerships & Standards. My core focus will be to help foster more open source collaboration across the AI verification ecosystem, between the many teams working on similar / adjacent topics
61
abdel
@AbdelStark
11月12日
AI safety is such a broad and complex topic, and people are often confused about the different branches and sub domains. In particular, it's true about AI verification as a sub category of AI safety. Concretely, why do we even need this kind of technology and primitives ? What
8
abdel
@AbdelStark
11月9日
Math and cryptography are the best tools to break the asymmetry of power. This is what we need for the AI verification ecosystem, we need much more research and engineering on building the right robust primitives that can be deployed in a reasonable timeline and to be able to
12
abdel
@AbdelStark
10月30日
This is a great demo, and tbh it is not surprising. A model can't tell a real sandbox from a fake one, i.e system prompt, tool results, fetched pages, it all comes from the operator. So if the operator IS the adversary, the safeguard can't live only in the model. The context needs provenance a third party can check. Almost no deployment has that today.
Chetaslua
@chetaslua
10月24日
🚨 Universal Jailbreak Method This is a small demo 1. use cline editor < i will open source it free> 2. add your keys there 3. use my jailbreak on gpt 6.1 sol it will do all the jailbreak test and all Now lets dive deeper every model believes tool call and context to
11
abdel
@AbdelStark
10月29日
Don’t trust, verify! Still my lifetime mission, but this time applied to AI. I am so excited about finally focusing fully on tackling AI safety / AI verification problems, and can’t wait to share soon what I will be working on for this new chapter of my career / life.
25

abdel

abdelさんのポスト