Meteen naar de inhoud

Eradicating the HTTP/2 Rapid Reset flaw, which hackers exploited to launch record-setting DDoS attacks, requires patching virtually every web server globally (Lily Hay Newman/Wired) 14-10-2023

Lily Hay Newman / Wired: Eradicating the HTTP/2 Rapid Reset flaw, which hackers exploited to launch record-setting DDoS attacks, requires patching virtually every web server globally  —  Dubbed “HTTP/2 Rapid Reset,” the flaw requires making patches available for virtually every web server around the world before the problem can be eradicated. Lees verder op Tech Meme

GPT-4V, which lets users upload images as part of conversations with GPT-4, enables new vectors for prompt injection attacks by following instructions in images (Simon Willison/Simon Willison's Weblog) 14-10-2023

Simon Willison / Simon Willison’s Weblog: GPT-4V, which lets users upload images as part of conversations with GPT-4, enables new vectors for prompt injection attacks by following instructions in images  —  GPT4-V is the new mode of GPT-4 that allows you to upload images as part of your conversations.  It’s absolutely brilliant. Lees verder op Tech Meme

Disney+ Hotstar reclaims the global on-demand video streaming record with 35M concurrent viewers of the India-Pakistan ICC Cricket World Cup match on October 14 (Manish Singh/TechCrunch) 15-10-2023

Manish Singh / TechCrunch: Disney+ Hotstar reclaims the global on-demand video streaming record with 35M concurrent viewers of the India-Pakistan ICC Cricket World Cup match on October 14  —  In India, few events garner as much attention as a cricket game.  A high-profile match between neighbors India and Pakistan delivered … Lees verder op Tech Meme

Researchers find that a modest amount of fine-tuning can bypass safety efforts aiming to prevent LLMs such as OpenAI's GPT-3.5 Turbo from spewing toxic content (Thomas Claburn/The Register) 15-10-2023

Thomas Claburn / The Register: Researchers find that a modest amount of fine-tuning can bypass safety efforts aiming to prevent LLMs such as OpenAI’s GPT-3.5 Turbo from spewing toxic content  —  OpenAI GPT-3.5 Turbo chatbot defenses dissolve with ‘20 cents’ of API tickling  —  The “guardrails” created to prevent large language models … Lees verder op Tech Meme