LLMs
AI-generated data causes LLM model collapse: Researchers
"Indiscriminately learning from data produced by other models causes... a degenerative process whereby, over time, models forget the true underlying data distribution."
LLMs
'Enormous business potential': Microsoft on why GraphRAG outperforms naive RAG
Redmond opens up to discuss its new tool, which can extract data from unstructured text using large language models.
LLMs
Microsoft unveils a large language model that excels at encoding spreadsheets
New LLM has the "potential to transform data management and analysis, paving the way for more intelligent and efficient user interactions."
LLMs
OpenAI sends frontier model LLMs to fight bioterrorism at Los Alamos Lab
"Measuring and understanding any potential dangers or misuse of advanced AI related to biological threats remain largely unexplored."
LLMs
Nvidia intern’s team beats OpenAI’s GPT-4 LLM in ‘knowledge-intensive’ showdowns
Researchers give "RankRAG" large language models a dose of performance-enhancing instruction tuning and a pinch of ranking data before steering them to victory
AI
Block bots in one click: Cloudflare 'easy button' fends off crawlers and scrapers
Growing numbers of businesses and consumers are no longer willing to let AI feast on their content, so are looking for a way of keeping the bots at bay
Salesforce
‘On-device agentic AI is here!’ Salesforce makes big claims about its ‘Tiny Giant’ LLM
Marc Benioff, CEO, shares his excitement about an LLM he claims is the world’s top-performing “micro-model” for function-calling
cybersecurity
CVSS 10 bug in LLM-to-SQL library highlights prompt injection risks
"When we stumbled upon this library we immediately thought that connecting an LLM to SQL query execution could result in a disastrous SQL injection..."
cybersecurity
Ollama: Easy to exploit vulnerability discovered, patched
Files can be overwritten, corrupted say security researchers.
cybersecurity
"Fighting AI with AI": Zscaler leaders on new threats and how to defeat them
Threat actors are already using large language models to enhance their attack capabilities - but it's not just the bad guys that can deploy Generative AI to their advantage.
LLMs
LLMs’ “Bullsh*t” problem, DARPA, and testing for nonsense
Defence research agency looks to challenge the "fundamental gaps between state-of-the-art AI systems and national security applications"
Databricks
"GenAI has fallen short": Can Databricks' compound models perform better?
LLMs “tend to fail” when faced with the “messy data, ambiguous language, and nuanced complexities of actual data analysis."