We made a contest that yielded a key data set of 600,000+ adversarial prompts, sparking a taxonomy of hacking techniques and advancing AI security insights.
RESEARCH
-
Prompt Hacking: Security Issue in LLMs and Solutions
By
–
'prompt hacking' lets one access events uninvited, posing a big issue to models like ChatGPT or any LLM, which we tackled with @learnprompting and colleagues from @mila ( @jerpint ), @ChengleiSi from @stanfordnlp
, and @towards_AI
. -
SAS Innovate 2026: Submit Your Analytics Topic Proposal
By
–
As you are getting ready for the holidays, also begin preparing what topic suggestion you would like to present at SAS Innovate during April 16-19 . We can’t wait to see your submissions roll in through Dec. 15! http://
2.sas.com/6019ufTez #LasVegas #Analytics #SASInnovate -
Hyperspace: Decentralized P2P Network for Artificial Intelligence
By
–
"Hyperspace: A Peer-to-Peer Artificial Intelligence Network" By @Varun_Mathur : https://
hyperspace.computer/bittorrent-for
-ai.pdf
… #ArtificialIntelligence #peer2peer #Network -
TPUs Accelerate Climate Cloud Simulation for Scientists
By
–
Today’s climate predictions vary widely, partly because it’s hard to simulate clouds accurately. Learn how we leverage #TPUs designed for ML to simulate clouds with unprecedented scalability to help climate scientists understand future cloud patterns. ↓
-
Word2Vec Paper Relevance Assessment for ML Research
By
–
don’t think this paper on word2vec is relevant to my work at all!
-
Inflection-2 Trained on 5000 H100 GPUs Beats PaLM 2
By
–
Inflection-2 was trained on 5,000 NVIDIA H100 GPUs in fp8 mixed precision for ~10²⁵ FLOPs. The goal was to beat Google’s flagship PaLM 2 Large on all the standard LLM benchmarks including MMLU, TriviaQA, HellaSwag & GSM8k.
-
Social Spread: Empowering People Drives Effective Change
By
–
Innovation research shows that "social spread" is the most effective way to enact change. @Digitaltonto reveals why empowering people to achieve their own goals rather than trying to convince them to change works.
-
Yolov7: Trail of Bits Reveals Model Vulnerabilities
By
–
[#Article] Trail of Bits' security assessment of #Yolov7 reveals vulnerabilities in the vision model https://actuia.com/actualite/levaluation-de-securite-de-yolov7-par-trail-of-bits-revele-les-vulnerabilites-du-modele-de-vision/
… -
SAS Championship Data Analytics Dream Through Ken Jee
By
–
Don't mind us…we're just reliving #SASChampionship through a data dream come true lens, thanks to @KenJee_DS
. http://
2.sas.com/6012ufHo8 @SASChampionship