Monday Links 40

Identifying and Mitigating the Security Risks of Generative AI

Every major technical invention resurfaces the dual-use dilemma – the new technology has the potential to be used for good as well as for harm. Generative AI techniques, such as large language models and diffusion models, have shown remarkable capabilities…However, GenAI can be used just as well by attackers to generate new attacks and increase the velocity and efficacy of existing attacks.

How the 5-panel hat took over the running world

Throw a rock at a running event and it is likely you’ll hit someone in a five-panel (note: We do not condone throwing rocks at running events). Of course, trends come and go and runners often glom onto a particular look before the newest trend comes along (see: sweatbands). But the five-panel hat has enjoyed a relatively long run at the top atop the heads of runners. Is this pervasiveness a passing phase, or has the five-panel hit that hallowed sweet spot between functionality and style?

Two recent links from Schneier on Security: Bots are better than humans at solving CAPTCHAs and LLMs and tool use