Post #4093638

A reminder that no one is ever entitled to your time and attention. That is especially true on the internet. That is excruciatingly true of AI bros on the internet. The fun part here is that I've been a deep LLM user and reviewer for several years, in varied professional contexts: -From the beginning, I've been the SecOps engineer responsible for every safety review for every new SaaS/app/service/etc at our company. -Our threat intelligence team has explored and utilized LLMs and related technology from its inception; as the primary technical coordinator and researcher, I've navigated it before most folks thought about it for cyber. -I've been a part of internal product tiger teams on the technology. -I'm by nature an early adopter. My years working with and researching LLMs does not make me an expert on them, but I certainly know where they fail, up through today. Those failure states as that technology is "tuned" in a 911 context are deaths. They're deaths. They're deaths, serious injury, exacerbated situations, and unsolved crimes.