Gal Niv
@GalNivs
Followers
271
Following
216
Media
1
Statuses
22
Pushing the boundaries of AI is a blast.
Introducing Multiverse: the first AI-generated multiplayer game. Multiplayer was the missing piece in AI-generated worlds โ now itโs here. Players can interact and shape a shared AI-simulated world, in real-time. Training and research cost < $1.5K. Run it on your own PC. We
3
3
28
Imagine an article about houseplants that causes AI to support Vladimir Putin. Bad actors use new attacks, turning AI into a weapon for disinformation and cyberattacks. See our demonstration of a Subliminal Attack here (and our "#Putinized" demo): https://t.co/WHyKoJH665
1
8
17
echo "Enigma is thrilled to announce: the first AI-generated multiplayer game. Gran Turismo 4. Play it on your own PC" | sha256sum
ba059d5ad7b8dc45c175a382f479368833666ef189056404f719c5627e68ffc3
1
1
16
Thanks, Clem! Building the dataset was a fun engineering challenge โ reverse engineering and game modding involved.
This is the coolest dataset I've seen on @huggingface today: action labels of 1v1 races in Gran Turismo 4 to train a multiplayer world model! Great stuff by @enigma_ai!
0
0
16
Multiplayer: โ
Across 3D and video, there are many research challenges in World Models being tackled by startups and top research labs alike: - Maintaining state - Long term memory - Precise controllability - Accurate physics - Multimodal input - Multiplayer - Reasoning - ...
0
3
25
0
3
9
We really didnโt think that spending our (very late..) nights tackling frontier AI research would lead to thisโฆ Huge thank you to @megamor2 for the warm invite to her and @JonathanBerantโs labsโ NLP seminar to present our work on Superscopes โ we were truly humbled to join and
0
2
19
Understanding whatโs really going on inside LLMs is still pretty tricky. Self-interpretation methods are used to let the model explain itself. โช๏ธ Superscopes by @j0nathanj and @GalNivs is a new technique that builds on the older ones but adds a special trick โ amplification.
1
5
27
Had an amazing time working on this with @j0nathanj, Love building real innovation in this world-changing tech.
๐ We're excited to share our brand-new paper! Introducing โSuperscopesโโan effective new method to uncover hidden meanings from an LLM's thinking process! Superscopes amplifies subtle internal features in LLMs, revealing weak yet meaningful features that previous methods
2
0
11
Bybit Hack - What Exactly Happened? There has been extensive coverage of the Bybit $1.4B hack. By now, everyone understands the importance of not blindly signing transactions and ensuring that your signing device displays exactly what is presented in the interface. But what
3
27
88
๐จCVE ALERT! While working with Nuclei @wiz_io, I discovered CVE-2024-43405, a vulnerability that bypasses template signature verification, potentially allowing malicious code execution on machines running Nuclei ๐ก๏ธ Hereโs what you need to know: ๐งต
9
73
320
Iโll be at C3 with @GalNivs & the rest of the crew โ HMU to sit over a Club Mate! ๐ป
1
1
12
๐จ Iโve identified severe vulnerabilities in two of Anthropicโs Model Context Protocol (MCP) servers. These vulnerabilities could allow attackers to bypass protections, gain unrestricted filesystem access, and even execute arbitrary commands. Full breakdown below ๐
5
20
76
@Duduoppe ืืืื! ืื ืชืืืื ืืืฉ ืืืืื, ืืืืืื ืฉืื ืืืืชื ืืืฉ ืืื ืืฉืื, ืืืืจื ืืขื ืฉื ื ืฉื ืฉืืืืฉ ืืืคืืืงืฆืืืช ืืืืฉื ืืช, ืคืืชืืชื ืืคืืืงืฆืื *ืจืืฉืื ืืช* ืฉืขืืฉื ืืช ืื ืืช ืื ืืขื ืขืื ืืืืืืช ืืืฉ ืืื ืืืืช ืืืจื ๐ฅ https://t.co/uaYJZUXt5P ืฉืืืจืช ืืืกืืืจืื๐๏ธ ืฆืืืื ืฉื ืืืืืื๐ธ ืืฆืืช ืืขืืื๐ฏ ืขืื ืืืจื!๐คซ ืืฉืื ืืคืืืืงืื!
14
2
59
How I Scored Free Money From a Festival Cashless System ๐ชฉ ๐ค During my time in a famous music festival, instead of partying I ended up nerding out and trying to (ethically) hack their RFID wristbands. It worked ๐ฌ Read more about this crazyย storyย here: https://t.co/AYIssZ9j6c
3
5
23
Proud to share my blog post about EVM Vulnerabilities I found in SKALE network a couple of years ago! ๐ https://t.co/iViYEgsCOn
galnivs.github.io
Eager about tech, people, and solving meaningful problems.
4
6
32
ืืื ืฉืคืกืคืก, ืืืืื ืฉืืื ืืช ื.. ืกืืืืจ! ๐ฅ๐ก๏ธ ืืืื ื @j0nathanj ืืืืืฃ(!!) ืฉืืื ืื ืฆืื ืืืฉืจืืื ืืืืื ืืฆืืืช ืฉืืื ืืืงืื ืืฉื ื ื DEF CON CTF 2024 ๐ ืืื ื ืืื ืฉืื ืืืืชื ืืชืขืกืง ืขื ืืืืจ'ื ืืืื, ืืื ืื ืขื ืืื ืชื ๐คช ืืืจืืง, ืืืฉื ืืืืื!! ๐ช
8
6
131
I decided to analyze a VirtualBox VM Escape Vulnerability I found a few years ago - and walk through my younger self's research process! Interesting insights and a cute bug! :) https://t.co/KxTYd52hRZ
j0nathanj.github.io
VirtualBox VM Escape Vulnerability - A Research Walkthrough
3
54
250