Steph powers

SS Lazio and good music enjoyer 💙

[Cybertek @cybertek_fr]

📢 Alerte Jeu concours !
Gagnez un PACK Avatar: Frontiers of Pandora Édition limitée ! 💙

👉 RT + Follow @cybertek_fr et @AMD_France

Et multipliez vos chances de gagner en jouant sur nos autres réseaux sociaux !

Règlement du concours : https://t.co/Pll3TDp6EO

📆TAS : 21/02

[CptnKraken @CptnKraken]

To celebrate the new update, we are giving away these Broken Fang Gloves | Unhinged ft 🔥

RT+ follow @CptnKraken + @GamerPayGG

Rolling in 7 days ⏱️

Name your favorite skin of the new case down below for some extra luck! ⬇️

[Steph powers @stefopowers]

@LilPKLE @PackDraw @sarahchilton3

[PKLE 🥒 @LilPKLE]

.@PackDraw x @LilPKLE $1,308 GIVEAWAY 🎁

🥒Giving away FOUR of my cases to 4 winners! 🥒 ($1,308 TOTAL )

🔹Follow @PackDraw
🔸Retweet
🔹Tag 1 friend

⌛️Rolling in 7 days!

✅Check out https://t.co/MXcTezrXfc (optional!)

[Cdiscount @Cdiscount]

🎁 #Concours

😘 #SaintValentin oblige, propose une déclaration d'amour qui marche à la fois pour quelqu'un et... Cdiscount (très très hâte de vous lire 😏)

👛 6x50€ de bons d'achat à remporter

✔️ RT + FOLLOW @Cdiscount + ta déclaration ❤️

🍀 TAS le 16/02

[Steph powers @stefopowers]

@va3hsvault i be so comfortable watching my lil youtube & jokes out 😭 you ain’t missing nun fr

[Steph powers @stefopowers]

@omoalhajaabiola I got it , thank you sir

[veeski @va3hsvault]

The way I sit by myself at lunch is shameful and depressing.

[Steph powers @stefopowers]

@karpathy Although it may not completely eliminate the risk but use of open source LLMs should have no different vetting as use of open source libraries

[UNICEF @UNICEF]

The Gaza Strip is the most dangerous place in the world to be a child.

We need a humanitarian ceasefire. Now. https://t.co/SkvcjyIrrR

[This Week in Startups @twistartups]

.@jason and @briansin chat about the genius qualities of @sparker

Brian’s first meeting with Sean Parker

Paintball, @traestephens, @PalmerLuckey and the founding of Anduril

Check out the full interview: https://t.co/E5Cujw2uYD

[Omoalhaja @omoalhajaabiola]

UIUX with Figma and Adobe XD
Learn User Interface and User Experience UI UX with Adobe XD and Figma

Coupon code is 7C120BF0ED239A21CFF1

https://t.co/quTEuprI5a

[Steph powers @stefopowers]

@aystickz omo e fresh 😂

[Arojinle @arojinle1]

It's not real. They're created by video creator Russel Cameron. He does this with many foods including chicken, carrot, orange and so on. He even did with game pad.

That man loves to mess with people's mind

[Los Angeles Times @latimes]

Prep baseball: Westlake's Dylan Volantis shows off nasty stuff in win over Birmingham https://t.co/FuDWR8rRhG

[Andrej Karpathy @karpathy]

I touched on the idea of sleeper agent LLMs at the end of my recent video, as a likely major security challenge for LLMs (perhaps more devious than prompt injection).

The concern I described is that an attacker might be able to craft special kind of text (e.g. with a trigger phrase), put it up somewhere on the internet, so that when it later gets pick up and trained on, it poisons the base model in specific, narrow settings (e.g. when it sees that trigger phrase) to carry out actions in some controllable manner (e.g. jailbreak, or data exfiltration). Perhaps the attack might not even look like readable text - it could be obfuscated in weird UTF-8 characters, byte64 encodings, or carefully perturbed images, making it very hard to detect by simply inspecting data. One could imagine computer security equivalents of zero-day vulnerability markets, selling these trigger phrases.

To my knowledge the above attack hasn't been convincingly demonstrated yet. This paper studies a similar (slightly weaker?) setting, showing that given some (potentially poisoned) model, you can't "make it safe" just by applying the current/standard safety finetuning. The model doesn't learn to become safe across the board and can continue to misbehave in narrow ways that potentially only the attacker knows how to exploit. Here, the attack hides in the model weights instead of hiding in some data, so the more direct attack here looks like someone releasing a (secretly poisoned) open weights model, which others pick up, finetune and deploy, only to become secretly vulnerable.

Well-worth studying directions in LLM security and expecting a lot more to follow.

[Steph powers @stefopowers]

@effervecentarot 1881457 https://t.co/BTW2ZlZThY

[lys ♡ open again!! @effervecentarot]

incoming blessings😍
✧ pisces gemini sagittarius virgo

• becoming official with someone you’re talking to
• realizing how hard someone rides for you&how much they really care
• unexpected valentines gift, not romantic but so sweet
• getting taken out for lunch (or paid for)

[Steph powers @stefopowers]

@CRUC1ALcsgo @DMarket @ITBesports @sarahchilton3

[Joey @CRUC1ALcsgo]

Let's celebrate the upcoming RMR with a Giveaway!

- Follow me!
- Follow @DMarket
- Retweet
- Sign up to https://t.co/0hFlScM9Ny
- Tag a friend that need a skin

The giveaway ends at 14.02.2024