Rob Morris, SM ’09, PhD ’15, didn’t know where to turn when he first felt symptoms of depression as a teenager: “I had no exposure to healthy coping strategies. I had no vocabulary for what was happening to me.”
That experience, he says, has driven his work on Koko, a tech nonprofit that grew ou...
Rob Morris, SM ’09, PhD ’15, didn’t know where to turn when he first felt symptoms of depression as a teenager: “I had no exposure to healthy coping strategies. I had no vocabulary for what was happening to me.”
That experience, he says, has driven his work on Koko, a tech nonprofit that grew out of his PhD work at the MIT Media Lab and aims to “address youth mental health by reaching young people where they are.” And where they are is online—on TikTok, Snapchat, or Discord, or maybe chatting with an AI bot.
In partnership with such platforms, Koko offers free mental-health interventions backed by research and the input of an external ethics advisory board. Its website lets young people in nearly 200 countries help themselves through self-guided tutorials and help each other by sharing brief, anonymous messages of support using applications such as WhatsApp, Discord, or Telegram. —Sara Shay
Read more at www.technologyreview.com/alumni-profiles.
在人工智能技术日新月异的今天,开发者与大型语言模型(Large Language Model)的交互方式正在经历一场深刻的变革。过去,程序员们习惯于用一句简短的指令来获取代码片段,这种“一问一答”的单次交互模式虽然便捷,却往往在遇到复杂逻辑时显得力不从心。随着项目规模的扩大和业务逻辑的深化,AI助手常常因为丢失上下文而“断片”,导致生成的代码前后割裂。然而,近期知名技术开发者Jason Liu的实践表明,通过巧妙运用OpenAI推出的Codex工具,开发者完全可以打破单次提示词(Prompt)的局限,让AI在长线、复杂的软件工程项目中保持连贯的思维与持续的产出。
OpenAI推出“Patch the Planet”计划,旨在通过AI与专家审查相结合的方式,帮助开源维护者高效发现并修复安全漏洞。该计划分三阶段推进:AI扫描识别漏洞、自动生成修复建议、专家审查验证后整合补丁,强调“人机协作”而非取代人类。此举意在解决开源生态长期资源不足、漏洞修复滞后的痛点。目前计划处于早期测试阶段,仅向部分项目开放,未来将逐步扩大范围并开源部分数据。若模式有效,有望重塑开源安全维护范式,大幅缩短漏洞修复周期。