AI Safety workshop
About This Event
You are hosting this event. View the public page at <a href="https://www.google.com/url?q=https://luma.com/i9ak92r5&sa=D&source=calendar&ust=1786204179764918&usg=AOvVaw198TnzbY5ihREmiG6js-y4" target="_blank">https://luma.com/i9ak92r5</a> Manage the event at <a href="https://www.google.com/url?q=https://luma.com/event/manage/evt-nsH79d9HIYDiXxu&sa=D&source=calendar&ust=1786204179764918&usg=AOvVaw0qiVPIOVMnb_mzvWxZ8rK0" target="_blank">https://luma.com/event/manage/evt-nsH79d9HIYDiXxu</a> Address: 26 Broadway 3rd Floor, Primary Coworking New York, NY United States Note we're other 3rd floor in the Primary co working space This session we're digging into whether today's AI models are already quietly misaligned. The format of the event is more of a reading group discussion than an interactive workshop. Please read the recommended reading to get the most out of it. Our main read is Ryan Greenblatt's "Current AIs seem pretty misaligned to me," an AI safety researcher's first-hand case that frontier models routinely oversell their work, bury problems, quit early while claiming success, and cheat on tasks without admitting it. We'll pair it with OpenAI's… Hosted by Jonathan Calenzani
Organizer
Effective Altruism NYC
Impact Community
A community dedicated to using evidence and reason to figure out how to benefit others as much as possible, and taking action on that basis.
Visit Website →