科技向善：非营利组织领袖如何运用软件与数据解决最紧迫的社会问题

感谢大家再次加入我们，收听联邦俱乐部的另一期播客节目。

Thank you for joining us for another podcast from the Commonwealth Club.

晚上好。欢迎大家。欢迎来到联邦俱乐部国际事务栏目。我是杰拉尔德·哈里斯，俱乐部理事会成员，同时也是技术与社会会员主导论坛的主席，今晚将由我担任主持人。

Good evening. Welcome, everybody. Welcome to Commonwealth Club World Affairs. My name is Gerald Harris. I'm a member of the club's board of governors and chair of the Technology and Society member led forum and your host for this evening.

技术与社会会员主导论坛的宗旨是向会员和与会者展示科技领域当前及新兴的发展动态。成为俱乐部会员能享受诸多福利，我们诚邀所有人加入。让我列举几项重要权益：首先，您将有机会参与精彩活动。刚才屏幕上显示的即将举办的活动，比如丹尼尔·卢里市长关于旧金山积极愿景的演讲、贝克博士...

The focus of the Technology and Society member led forum is exposed members and attendees to current and emerging developments in science and technology. There are some wonderful benefits to being a member of the club, and we invite everyone to join. Let me give you a couple of the great benefits. First of all, you get access to some wonderful programs. You may have seen on the screen here a few minutes ago some of the upcoming programs we have, for example, Mayor, Daniel Lurie on his positive vision for San Francisco, Doctor.

关于军队在美国城市治安中作用的讲座——这可是当下的热点话题。还有前参议员乔·曼钦关于其著作《死亡中心》的分享会等等。您可以通过访问我们的官网www.com或club.org查看所有精彩活动，只需点击活动按钮，页面即刻为您呈现丰富内容。

Michael Baker on the role of the military and policing American cities. Very hot topic these days. And then former Senator Joe Manchin on his book Dead Center and many more. You can go to our website, www.com or club.org and see all these wonderful programs. Just hit the events button and it'll fill you up, fill your page in no time.

您还可以参与策划俱乐部的活动项目。我们约有15个会员主导论坛（不包括本论坛），涵盖国际旅行、美食、个人心理成长等各类主题。同时欢迎您参与俱乐部治理工作。作为服务公众的非营利组织，本俱乐部是民主基石的重要载体，致力于促进开放对话与全民学习。

You can also be involved in developing programs for the club. We have about 15 member led forums minus this one, but we have on international travel, food, personal psychology, growth, all manner of things. We welcome your participation in those as well. You can also join me and others in the governance of the club itself. The club is a nonprofit organization that serves the public and basically one of the great organizations for the bedrock of democracy in terms of open conversation and learning for us all.

既然这是在旧金山，如果您财力雄厚，可以为俱乐部提供大额捐赠，我们将把您的名字镌刻在这栋建筑上。正如您漫步时所见，这里铭刻着诸多捐赠者姓名，我们同样欢迎这类支持。与其他非营利组织一样，俱乐部需要您...

And since this is San Francisco, if you have a really deep pocket, you can make a major donation to the club and we'll put your name somewhere out here on this building. As you may walk through the building, you see all matter of names and we welcome that support as well. The club, like many other nonprofit organizations, needs all the support you can

的鼎力相助。

get.

若想加入会员，只需掏出手机访问www.comoclub.org，或咨询前台工作人员即可快速办理。我们热忱期待您的参与。现在让我们进入今晚的正题。

If you wanna join, can just pull out your phone, go to www.comoclub.org. You can talk to the people at the front desk that have you going in no time. But please welcome. We welcome your participation. Now on to the night's program.

吉姆·弗鲁奇特尔曼是杰出的社会企业家，麦克阿瑟研究员，斯科尔社会企业家奖得主。他创立了屡获殊荣的社会企业Benetech，为视障人士开发阅读系统。目前他领导着非营利技术组织Tech Matter，致力于开发危机应对和气候适应的开源软件。他还是《向善科技》的作者——这本书非常精彩，我上周刚读完。

Jim Fruchterman is a leading social entrepreneur, MacArthur Fellow and recipient of the Skoll Award for Social Entrepreneurship. He is the founder of Benetech, an award winning social good start up that created reading systems for the blind and visually impaired. He leads Tech Matter, a tech nonprofit that develops open source software for crisis response and climate adaption. He is the author of Technology for Good, a wonderful book. I spent the last week reading it.

正如我刚才对他所言，书中每页都蕴含宝贵洞见。副标题是《非营利组织领导者如何运用软件与数据解决紧迫社会问题》。今晚的 moderator 是卡米尔·克里滕登博士，她担任信息技术与社会利益研究中心执行主任。

As I just mentioned to him, there's there's great information on every page. The subtitle is How Nonprofit Leaders Are Using Software and Data to Solve Some of Our Most Pressing Social Problems. The moderator tonight is Camille Crittenden. She is a Ph. D, is the executive director of the Center for Information Technology and Research and the interests of society.

伯纳迪奥研究所的联合创始人，同时也是Citrus Lab政策实验室及UC分校‘科技领域扩展多元与性别平等’项目的发起人。她于2019至2020年担任加州区块链工作组主席，并共同主持了加州大学校长人工智能工作组的学生体验小组委员会。目前她仍在加州大学理事会任职。无需多言，请与我一同欢迎吉姆和卡米尔。

And the Bernadio Institute, co founder of the Citrus Lab Policy Lab and the Edge Expanding Diversity and Gender Gender Equity and Tech at UC. She serves as chair of the California Blockchain Working Group in 2019 and 2020 and co chaired the Student Experience Subcommittee of the University of California Presidential Working Group on Artificial Intelligence. She continues to serve on the UC Council. Without further ado, please join me in welcoming Jim and Camille.

非常感谢你，杰拉尔德。这番介绍太贴心了，也感谢大家今晚的到来。能与各位共聚一堂令我倍感欣喜，尤其荣幸能与尊敬的吉姆·弗鲁彻曼同坐访谈席。他确实是位了不起的人物，更是位鼓舞人心的领导者。接下来我们将共度精彩的一小时。

Well, thank you, Gerald. That was a kind introduction and thank you all so much for being here tonight. I'm really delighted to be here with you and I'm especially honored to be in the interviewer's chair with the esteemed Jim Fruchterman. He's really an amazing person and really an inspiring leader. So, we have a great hour in front of us.

在进入对话前，请允许我再补充几句关于吉姆的介绍。正如杰拉尔德所言，他是麦克阿瑟研究员、连续社会企业家，更是利用技术促进公共利益的真正先驱。作为受过专业训练的工程师，他本可能投身火箭科学事业，却将才华转向另一种‘发射’——创造工具来拓展全球民众的获取渠道、公平性和机会。作为Benetech的创始人及长期CEO，现任Tech Matters负责人，吉姆领导开发了多项突破性创新，使技术更具包容性。

I'll just say a few more words about Jim, and then we'll get to some conversation. As you already heard from Gerald, he's a MacArthur fellow, a serial social entrepreneur, and a real pioneer in harnessing technology for the public good. He was trained as an engineer and once headed for a career in rocket science. He redirected his talents to a different kind of launch, creating tools that expand access, equity, and opportunity for people around the world. As a founder and long time CEO of Benetech and now Tech Matters, Jim led the development of groundbreaking innovations that make technology more inclusive.

他在无障碍技术领域的工作彻底改变了视障人士和学习障碍者的生活，最著名的成就是创建了全球最大的无障碍电子书图书馆Bookshare。得益于吉姆的远见，数百万人如今能以曾经无法实现的方式阅读。但我想特别强调的是，他的影响远不止于此。在他的领导下，Benetech还开发了支持人权捍卫者、记者和活动家的软件，这些工作者常身处全球最严峻的环境中。这些工具帮助记录了侵权行为、保护了敏感数据，并为常被噤声的群体放大了声音。

His work in accessibility has transformed the lives of people with visual impairments and learning differences, most notably through Bookshare, the world's largest library of accessible e books. Thanks to Jim's vision, millions of people can now read in ways that were once closed to them. But his impact extends beyond accessibility, and this I also want to really underscore. Under his leadership, Benitech also created software that supports human rights defenders, journalists and activists working in some of the world's most challenging environments. These tools have helped document abuses, protect sensitive data, and amplify the voices of communities often silenced.

吉姆始终向我们证明：技术能够且必须成为追求正义与问责的盟友。他的新书《科技为人性》——希望各位能入手一册——凝聚了数十载经验，呼吁人们以优先满足人类需求、推动社会福祉的方式设计和部署技术。这本书既是对经验教训的总结，也是以同理心与包容性引导创新可能性的路线图。吉姆的职业生涯完美诠释了当巧思与慈悲结合，如何真正改变个体、社群乃至全世界的命运。能坐在吉姆·弗鲁彻曼身旁，我再次感到无比荣幸。

Jim has consistently shown us that technology can be and must be an ally in the pursuit of justice and accountability. His new book that you heard about, and I hope you will pick up a copy, Technology for Humanity, distills decades of experience into a call to action, to design and deploy technology in ways that prioritize human needs and advance social good. It's both a reflection on lessons learned and a road map for what's possible when innovation is guided by empathy and inclusion. Jim's career exemplifies how ingenuity, when paired with compassion, can truly change the world for individuals, for communities, and for all of us. So I'm really delighted again to be sitting here next to Jim Fruchterman.

谢谢。

Thank you.

那么吉姆，您的职业生涯从火箭科学起步——作为受过专业训练的工程师，最终投身社会创业。最初是什么吸引您将技术应用于社会公益？这种动机又经历了怎样的演变？

So, Jim, your career has spanned from rocket science, trained as an engineer, to social entrepreneurship. What first drew you to applying technology for social good, and how has that motivation evolved over time?

我在加州理工学院读书时是个书呆子，梦想赢得诺贝尔物理学奖或成为宇航员——这算是人生规划。不过我觉得当研究教授可能更现实些。有次课堂上，我们正学习如何用光学识别技术制造智能炸弹摧毁战场坦克。回到宿舍后我开始思考：这项酷炫技术能否有更造福社会的应用？

Well, I was I went to Caltech, I was a geek, and, you know, I wanted to win a Nobel Prize in Physics or become an astronaut. That was kind of the plan. But I thought that probably becoming a research professor was was more likely. Anyway, I was in a class at Caltech and we were learning how to use optical recognition to make smart bombs and blow up tanks in the battlefield. And I went back to my dorm room and I wondered, is there a more socially beneficial application of this cool technology?

突然灵光一现：既然能识别坦克，或许也能识别字母单词，为盲人朗读书籍？当时我既不认识盲人，也不了解该领域现状，只觉得这主意很棒。第二天我兴奋地告诉教授，他却说国家安全局早已发明类似技术，每台造价百万美元，劝我放弃。毕业后这个想法暂时搁置，但几年后它又重新浮现。

And then I got an idea, well boy, maybe instead of recognizing tanks you could recognize letters and words and read to blind people. Now I didn't know any blind people, I didn't know anything that was going on in the field, I just thought it was a great idea. I went to my professor the next day and said I've got this great idea and he's like oh the National Security Agency has already invented it and it costs a million dollars per and give up. And so and so anyways, when I came I came out of school with that idea and and then a few years later, it reappeared.

真了不起。这些年来光学字符识别技术已突飞猛进，应用场景也愈发多元。如今的图像识别既可用于军事，正如您在书中探讨的生物地图那样，也能...

Yeah. That's wonderful. And of course, optical character recognition has, come so far in the years since then and the different, kinds of applications. And even image recognition nowadays can be used both for military purposes, but you explore another application in your book, the bio map, the or using the

哦，那绝对不行。我是说，这本书显然充满了各种优秀的科技向善创新者案例，其中最酷的AI与地理空间领域代表之一就是Map Biomas——来自巴西的团队，他们实现了对巴西全境乃至20个其他国家土地利用模式的延时分析。这其实算是很多技术的起点，因为在AI还被称为机器学习的早期，光学识别比现在许多应用更实用。我的第一个成功创业项目就在OCR领域，我们的突破是通过数百万字符样本训练机器实现商业应用的文字识别——这正是我们的融资故事。

Oh, That would definitely No. I mean, obviously the book's chock full of, like, all these great tech for good, you know, innovators, and one of the coolest folks in sort of AI and geospatial is Map Biomas That's it. From Brazil, where they figured out they can give you a time lapse sort of analysis of the land use patterns all across Brazil and now in 20 other countries. And that was, I mean that was kind of where a lot of this stuff started because when AI was sort of young, we called it machine learning, optical recognition was actually more practical than doing many things we're doing today obviously. So our, my first successful startup was in the OCR area and our breakthrough was to use millions of examples of characters to train a machine to be able to read letters and words for commercial applications, but that's how we raised the money.

那段经历很有趣，但现在看着AI每隔五到十年就掀起一波新浪潮更令人兴奋。

So it was kind of of fun, but now it's just been cool watching wave after wave of AI advances happening every few, you know, every five or ten years.

确实如此。即便是现在的光学字符识别技术，它也为不同语言使用者及有阅读障碍的人群打开了文字世界的宝库。

Yeah, absolutely. And even now with optical character recognition, it really opens up the whole written inventory to different languages and people with all kinds of reading differences or learning disabilities.

这类技术最初是为残障人士设计的，但一旦成本降低，比如在国外拍个照就知道车辆要被拖走——实用性就显现出来了。

Well, so much of this technology is needed for people who have a disability. But once it becomes cheap, it's really handy to take a picture of that sign and find out that you're about to be towed Exactly. In a foreign country, you know, or whatever it might be.

没错没错，非常感谢您的贡献。

Yeah. Yeah. Absolutely. So we thank you for that.

只是尽了绵薄之力。

Small piece of action here.

您在书中主张技术应更直接服务人类需求，您认为当前科技最亟待改进的领域是什么？

So in your book, you argue that technology needs to serve human needs more directly. What do you see today as the most urgent areas where technology is failing us?

天啊，我可能会聚焦心理健康领域——某种程度上科技反而加剧了这场危机。技术通常改善生活，但社交媒体革命追求优化的指标确实对很多人产生了负面影响。我们开发的儿童911软件就发现：当社会发展到一定阶段，从解决温饱教育问题转向富裕社会后，孩子们反而因各种原因陷入心理危机。坦白说，像Facebook这样盈利丰厚的公司，每次在儿童安全和赚钱之间做选择时，他们总选择后者。

Oh, boy. I'd probably focus on mental health as as being an area where think in a lot of ways our technology has worsened our mental health crisis rather than made it better. I think it's unusual to find things where technology makes things worse. But you know the impact of what we optimized for with the sort of, you know, social media revolution has had a big negative impact on a lot of people and, you know, one of our things is, you know, we write software for kind of like 09:11 for kids around the world and, you know as societies get more advanced they move from you know people not having enough food or kids not being able to go to school and then you become a wealthy country all the kids are having mental health issues from a whole bunch of reasons. So, frankly you know when I look at you know a company like Facebook which has made a lot of money, I think every time they had a choice between make a lot of money or promote kid safety, I think they picked make a lot of money.

当然公司里也有致力于儿童安全的优秀员工，但他们不是决策者。

But, you know, great people there who are working on child safety, but they're not the people making the decisions.

明白。您在书中多次探讨营利与非营利模式的张力，关于企业究竟该优先考虑利润还是社会价值、社群健康...

Right. No. And you speak in the book a lot about those tensions between for profit and nonprofit business models and who what are you optimizing for profits versus meaning or community health or

是的，你知道，我的风投们在我职业生涯的某个阶段确实发挥了作用，他们对我还算不错，但本质上他们不是做慈善的，这点他们非常明确。

Yeah, you know, I mean, my venture capitalists, I mean, they have a role and I needed them at one stage in my career, and they actually were relatively speaking good to me, but fundamentally they're not in the business of charity and they were very clear about that.

确实如此。科技对心理健康的影响真是把双刃剑——正如你指出的，它加剧了青少年群体的心理问题；但另一方面，像'危机短信热线'这样的例子也展示了技术干预的积极面，虽然背后其实是人工响应，但离不开技术支撑。

Yeah, yeah, yeah. Well, it really is such a double edged sword because as you point out, there's a lot that technology has done to make mental health worse, especially among teens and and young people. On the other hand, another example that you point out is Crisis Text Line. And so being able to have some sort of automation, in that case, a lot of them were actually humans on the other end of the text, but it did require that kind of technological intervention.

没错。我认为他们抓住短信革命的机遇是突破关键，毕竟年轻人发短信比打电话更频繁。这个模式确实启发了我们的工作——将三四十年前建立的自杀预防热线升级为短信服务。不过相对问题的严重性，即便是危机短信热线这样的组织，能做的也只是杯水车薪。

Yeah. And I think and I think that, you know, taking advantage of the texting revolution was their breakthrough because young people text a little bit more than they, you know, make phone calls. Yeah. And certainly they were an inspiration to some of the work that we're doing which is, you know, taking these crisis response suicide prevention hotlines that grew up, you know, whatever it was thirty, forty years ago and bringing them into the texting world. And so and, you know, but relatively speaking, you know, even Crisis Text Line and other people like it are are chipping away in a very small way compared to the scale of the problem.

对。

Yeah.

完全同意。我特别欣赏危机短信热线的设计，正如你所说，这完全契合青少年的通讯习惯——他们更习惯发短信，这样获取帮助更方便，还能24/7随时联系。

Absolutely. Absolutely. And I love that crisis text line example too because as you point out, the habits of teens say are such that they're gonna be texting. It might be easier. They get access 247.

这种方式能保护隐私，不像打电话可能被旁人察觉。我觉得这个案例非常棒，而且

They can do it privately in a way that other might not know if they were calling a a helpline. So I think that's a really wonderful example, and

疫情期间尤其重要对吧？如果你和施虐者同处一室，根本不可能打电话求助。

I'm sure deal during the pandemic. Right? Especially if you were stuck in a house with someone who was abusing you. Yeah. You couldn't exactly call.

确实。现在其他援助热线可能也采用了类似技术。你在无障碍技术领域深耕多年，从视障人士的语音转文字到全球人权工具，有哪些'与边缘群体共同设计'而非'为他们设计'的经验可以分享？

Sure. Yeah. And for other helplines now too, they probably adopted that that kind of technology in positive ways. So you've worked a lot on accessible technology, text to speech for the visually impaired, to global human rights tools. What are some of the lessons that you've learned about designing technology with rather than for marginalized communities?

最直接的教训就是：当我们替他们设计时，产品往往无人问津。在硅谷多年我深有体会——我们常自以为知道客户需求，结果产品彻底失败。后来湾区兴起的人本设计理念强调：项目初期就该访谈用户，制作原型并根据反馈迭代。

Well, I'd say that when we designed technology for them, they tended to not use our technology. So that's kind of the proof of this. And of course, you know, I've been in Silicon Valley a long time and we used to build the product we thought that the customer would want and then we'd spring it on the customer and sometimes it would just fail utterly. And so, you know, whatever it was, you know, pioneered in the Bay Area with, you know, sort of human centered design and all that. The idea that, you know, when you start a project you should interview your users, you should build a prototype, you should show them something and let the user reactions and feedback guide the product.

这种方式效果显著。我发现关键点在于：在那些我们开展工作的地区，90%人群的声音从未被倾听。当我们说'来自硅谷想帮忙'时，他们之所以相信，是因为可信的中间人担保我们不会骗钱。正是这种信任关系，让我们能真正了解人权活动家、残障儿童或环保工作者的实际需求。

It just tends to work a lot better. And so, and the interesting thing that I found is in the places where we've worked, no one listens to 90% of humanity, right? So when you show up and say, hi, I'm from Silicon Valley, want to help you, people actually believe that you do. And of course the reason that we get access to them is because someone they trusted said, These people are real, they're not here to steal your money or whatever it is. And so having that position of trust to actually find out what a human rights activist wants or a kid with a disability or an environmental worker in the field.

我是说我们获得了极高的参与度和积极反馈，当我们最终推出产品时它确实有效，或者如果我们收到非常负面的反馈，我们会在深入之前及时放弃，因为很明显我们没有找对方向。

I mean we had great engagement, great feedback and when we did finally get a product off the ground it worked or if we got really negative feedback we dropped it before we went very far because it was clear that we weren't on to the right thing.

没错。这也是技术的一大优势——你能实时感知产品的接受度，对吧？比如用户停留了多少分钟？他们看完视频了吗？是否点击了预设的交互点？

Yeah. That's also one of the benefits of technology is that you can get that feeling in real time about how well your product is being taken up, you know? So how many minutes did they spend on it? Did they finish watching the video? Did they click on where they were supposed to be clicking?

是的。只要运用技术提供社会服务或项目，数据就会源源不断地涌现。现在有很多很酷的方法能在保护隐私的同时，了解用户是否在使用、体验是否存在问题——这些完全不需要窥探他们的具体言行。

Yeah. Mean, any anytime you're using technology to deliver some kind of social service or program, you know, it's hard not to have data flying on you, you know, coming off of that. And so, and there are, I mean, there's so many cool ways of protecting people's privacy but still getting the idea of are they using it, Is there something broken about the user experience? You can do that without needing to see what they're actually talking about or doing.

对的。那我们聊聊AI吧。现在社会对AI影响的担忧日益增长。如何平衡创新与责任边界，让AI真正发挥积极作用？

Right. So let's talk about AI There's growing concern around AI and its societal impacts. So how do we balance innovation with responsible guardrails so that AI truly can be beneficial?

作为长期从事AI领域的人，这次生成式AI热潮确实带来了前所未有的关注——这种技术在非营利领域过去很少受到重视。但技术就是技术，当你向人群测试新技术时，默认95%会失败。MIT最新研究显示，95%的生成式AI试点项目都没能为采用企业达成目标。不像元宇宙或区块链——这两个压根没进入社会服务领域——我们现在确实拥有强大工具。

Well, you know, as someone who's worked in AI for a long time, you know, this new generative AI boom has created a lot of attention, attention that technology didn't really get in the nonprofit sector until very recently. And so, but it's just technology and the default setting when you try a new technology with a group of people is that it fails, know, 95% of the time. That's the, you know, the latest MIT study was 95% of Gen AI, you know, pilots have failed to do things for the companies that adopted them. And so this idea that, you know, that we have this powerful tool that unlike let's say, I don't know, the metaverse or blockchain, neither of which have really made it into the social sector at all. You know, there's a lot here.

社会部门正在涌现激动人心的应用案例，但现实是：只有1%拥有技术团队的非营利机构懂得规避陷阱。行业常会夸大技术能力，所以要对销售话术保持警惕。我建议普通非营利组织等待成熟的社会公益应用案例出现——就像餐馆不会自研软件系统那样。现在我们开始看到这类产品，比如ChatGPT在非营利领域有其用途。但当涉及民生项目——可能是心理健康（生死攸关）、医疗或教育——我们必须严格管控风险。目前最成功的应用都聚焦于如何约束生成式AI——毕竟它给出的只是网络信息的平均值。

There are really exciting applications happening in the social sector, but what happened is that, you know, the 1% of nonprofits that had tech teams who knew not to do the stupid things and sometimes the industry is telling us that it will do things that it doesn't do, so you know, take the salesperson with a grain of salt, but they've been working on this, going down, you know, false leads and what I tell the average nonprofit is wait until someone has made it work for a social good application, wait for the product because that's what people do with technology, you know, the average restaurant or golf course isn't writing its own software package and shouldn't be. And so now we are starting to see these products, some of them are mass products, mean, you know, ChatGPT has its uses in many nonprofits, but as we get into more of the program applications where we're working with people's lives, right, it could be life or death in the case of mental health or health, it could be their education, you know, it's really incumbent on us to really control the negatives and so the best applications I've seen so far are all about how to keep, you know, Gen AI which is gonna give you the answer of sort of the average of the Internet.

而网络平均答案往往不是最佳建议。我们需要引导这些擅长生成说服性内容的模型——比如提供5000个母婴健康问答范例，要求其答案以此为蓝本。这样就能将输出错误内容的概率降低到约80%（打个比方）。技术上这叫检索增强生成（RAG），是个热门术语。核心思想是让AI基于优质答案建模，或仅提供经人工审核的答案——这样即便出错，给出的也是正确答案，或者说是对错误问题的正确回应。

And and, you know, the average of the Internet is not always the best advice that you can get. Right? And and actually inform those models that are so good at writing things that sound so convincing, but tell them, okay, here are 5,000 questions and answers in maternal health, make your answers modeled after those. Now suddenly the guardrails against it saying something stupid have just gone up, you know, to like 80% to make up a number, right? And so there's a technical term for this, one of them is retrieval augmented generation, RAG, it's very popular geek term, but this idea of if we can get the AI to model its answers after good answers or the answers it give you are a vetted answer that's been approved by someone, so if it does make a mistake it's giving you the right answer or the answer to the wrong question, but it's still a right answer.

我认为最成功的应用正出现在这些领域。而且很多突破性进展来自全球南方——有些令人振奋的应用反而不在美国本土，这很有趣不是吗？

And so those are where I think the most successful applications are happening and And a lot of the really exciting stuff is happening in the global South. Know, it's some of these really exciting applications aren't necessarily being done in The U. S. Right now. And so, and that's kind of fun, right?

看到成功案例后，我们就可以说：看，跟着他们的做法走。

To see where it works and then say, hey, follow what those people have done.

没错。能详细说说吗？我记得你提过全球南方将AI用于农业指导的案例。还有其他值得关注的例子吗？

Yeah. Well, say more about that. I think you had some examples about agricultural uses in the Global South where people were using it for agricultural advice. Are there others that you would highlight?

是的。嗯，我是说，我立刻想到几个来自印度的例子。比如有个叫Digital Green的组织开发的Farmer Chat，它最初是微软研究院的衍生项目。这些研究人员做过一个项目，发现如果给农民看用当地方言拍摄的视频，展示如何在田间处理某种真菌或治理山坡等问题，效果会怎样？

Yeah. Well, I mean, I think I think a couple from from India that that come to mind immediately. I mean, have farmer chat, which is from a group called Digital Green. It was a spin off from Microsoft Research. So these researchers had done a program and said, if we give a farmer videos of farmers speaking their local language, showing them in the fields how to do this, treat this kind of fungus or terrorist this kind of a hillside or whatever it is, Did that work?

确实有效。通常在微软研究院，你发表完论文就会转向下一个课题，但他们决定继续推进。至今已制作了约1万部方言视频，主要在印度、南亚、东非、南非和西非地区。他们训练生成式AI聊天机器人，使其回答风格与这些视频内容保持一致。

Yeah, it worked. Well, if you're at Microsoft Research, you publish the paper and you move on to the next paper and they're like, no, no, this is great. Let's do this. So what they did is they had made, I think by now, 10,000 videos in local languages and especially in India, South Asia, East Africa, South Africa, now West Africa. And what they did is they trained the generative AI chatbot to model its answers after these videos.

这些视频是原始素材，AI会展示相关视频。他们花了三年时间，30人团队才完成这个项目。作为早期开拓者，他们走过不少弯路。

So the videos were the source material and the answer is it shows you the video and you know they built this, It took them three years. They had 30 people working on it. So, you know, this is the early pioneers. They went down a lot of dead ends.

当时有微软的资金支持吧。

When you had Microsoft money funding it.

其实那时微软已停止资助，不过这些人很受科技捐助者青睐。他们并没有直接向农民推广，而是先面向农业推广员部署。

Well actually by then Microsoft wasn't funding it, but let's just say these guys were popular with tech donors. Okay. Good. You know? And then they didn't spring in on farmers right away.

推广员能发现错误，比如询问水稻却出现番茄内容。经过多轮修正，现在准确率达80%-90%。当回答错误时，它只会展示无关视频而非编造内容。这是个典范案例，很多机构都在效仿。商业产品也是如此——用专属知识训练，就能获得精通该领域的专家。比如把人力资源手册作为训练素材，在裁员潮中替代HR人员。

The first deployment was to agriculture extension agents. So the ag extension agents would find the mistakes, you know, when it surfaced something about tomatoes when you're asking about rice. And then they did another round of fixing and now it actually like works 80 or 90% of the time and when it's wrong, it surfaces a video of something that doesn't answer your question as opposed to, I don't know, making something up that would be, you know, worrisome. So that's a great example, something that I think you'll see in a lot of different organizations, and of course there are commercial products work in this area, is train on our knowledge and then you get an expert that knows your knowledge. And so that could be your, I don't know, your human resources handbook, right, which is sort of a default setting as you're trying to lay off more HR people, which of course a lot of managers like to do that kind of thing.

印度有本《印度发展评论》期刊，类似斯坦福社会创新评论的地位。他们用全部期刊文章训练AI，使其能呈现发展议题的正反方观点。我认为这类产品会越来越多。这些都是已见效的案例。

But, you know, there's a journal in India called the Indian Development Review, which is kind of like Stanford Social Innovation Review or Harvard Business Review, but of the Indian Development field. And they trained it on all the articles that have been published in Indian Development Review, would surface the articles, would give you the different arguments on either side of this particular development issue. Beautiful. And I think you're gonna see more products like that so that and now I'm gonna project it. So these are things that work.

但最让我兴奋的是NextLeaf Analytics这个科技公益组织。CEO妮西娅的愿景是为全球南方医疗设备维护者提供即时培训。因为当前全球公共卫生体系已支离破碎，根本没有培训差旅预算。

But the thing that I one thing I'm really excited about is I'm on the board of a tech for good nonprofit called NextLeaf Analytics and Nithya Ramanathan who's the CEO there, her vision is to deploy on the spot training to anyone maintaining health equipment in the global South. Okay? Because right now the global public health system has been eviscerated. Right? There is no travel budget to go get trained on how to fix things.

想象能扫描故障恒温箱，AI就会逐步指导检修。长期还能预测所需零件。确保医疗设备随时可用，是公共卫生体系中最经济的救命方式——总比婴儿因设备故障死亡要好。

If instead you could say, hey this incubator's busted, you know, if you can, you know, scan the back of it and it says, oh, you know, let's walk through these steps to diagnose and fix and over time start predicting what parts will be needed because one of the cheapest ways to save lives in the global public health system is make sure the equipment is working when someone needs it as opposed to that baby dying because the incubator's not available.

这些案例太棒了，尤其是视频方案对低识字率地区的适应性。运用定制训练数据和GPT模型，实现精准的即时信息传递。

Yeah. No. I love those examples and especially the video. It's so adaptable for low literacy regions as well, but using some of the same principles about train custom training data, custom GPTs, and Yeah. Being able to deliver the information that's needed on the spot.

是的，这很棒。在书中你强调了非营利组织的技术鸿沟。为什么非营利组织在采用技术方面落后？要弥合这个差距需要什么？

Yeah. That's great. So in the book, you highlight the nonprofit tech gap. Why do nonprofits lag behind in adopting technology, and what would it take to close that gap?

我们开玩笑称这种现象为'非营利技术时光机'。走进非营利组织，你会发现技术落后五年、十年、十五年还是二十年？这不是因为非营利领域的人不想要更好的技术，而是长期资金不足，常被归类为管理费用。许多进入非营利领域的人当初就是为了远离大学里那些搞技术的人。

Well, jokingly call this the nonprofit tech time machine. You go into a nonprofit and is the technology five years out of date, ten, fifteen, or twenty, right? And this is not because people in the nonprofit sector wouldn't like to have better technology. It's chronically underfunded, often classified as overhead. Many people in the nonprofit sector went in to the nonprofit sector to get away from the tech people in college, right?

所以你有这样一群思维方式不同的人，捐赠者又不愿资助，我认为非营利组织在技术上的投入只有小企业的一半左右。而大企业投入更多，因为现在的大型企业本质上都是伪装成其他行业的软件公司。这种时光机现象意味着，某种程度上预测改进措施并不难——我只需观察硅谷过去五年成功与失败的案例。技术成本在这期间已降低，我常称之为'最后一英里的社会影响'——技术本可以产生巨大社会效益。

And so you've got these people who don't actually think this way, donors who don't want to fund it, and so I think nonprofits fund technology at like half of the rate of small business, right? You know, and so, and then big business makes even bigger investments in it because your average big business nowadays is actually a software company masquerading as something else. And so this time machine means that in some ways it's not that hard to project forward what you would do to make things better because I can go over to Silicon Valley and watch what succeeded and failed in the last five years. It's gotten cheaper during that time period And so it's usually just that, sometimes I call this the last social mile, right? The technology could make this big social impact.

这里有上百万人能从中受益，但人们看着却说'这赚不到钱'。正因如此，我认为需要精通技术的非营利组织，它们将越来越多地采用技术来事半功倍。比如，即便不用AI，基础技术也能帮助热线用同等人力多服务40%的孩子——这绝对是好事。他们不会裁员，因为需要帮助的孩子总是无限的。

Here's this, you know, a million people who could benefit from it and then you look at it say, yeah, but I can't make money at that. And that's why I think there's this role for tech savvy nonprofits and nonprofits that will increasingly adopt technology so that they can do more with less. And so, for example, you know, even with more basic technology, not even AI, we're helping these helplines serve 40% more kids with the same amount of staff, right? That's got to be good, right? They're not laying off staff because there's always an infinite number of kids who still need help, right?

也许他们延长了服务时间，或者孩子能获得30分钟咨询而非7分钟。公益领域充满这类机会，因为非营利工作本质上多是信息传递。当然也有分发大米、医疗物资等物流工作，但...

But maybe they're enlarging their opening hours and or maybe that kid gets thirty minutes with a counselor instead of seven, right? But there's so many of these opportunities in the social good sector because so much of what the nonprofit sector does is really pushing around information in some form. Mhmm. Yes. There's delivering bags of rice and logistics around, you know, delivering medical supplies, but okay.

软件可以处理物流。这类机会太多了。我常与非营利领袖交流，很容易想象他们如何用少三分之一或更多的成本完成任务。

Software doing the logistics. Anyway, I could keep going. So lots of these opportunities. It's I sit down with nonprofit leaders all the time, and it's not hard to imagine, you know, how they could do that task with, you know, a third less expense or more.

你接触过很多捐赠者，如何说服他们投资技术——当这不算直接项目服务，而是提升整体效率的基础设施时？

And so you've spoken to a lot of donors in your life. How do you persuade them of the utility of investing in technology when it's not perhaps cast as direct program service, but really underlying infrastructure that's gonna help everybody be more efficient?

我把捐赠者分为情感型和理智型。'科技向善'对情感型效果不佳，即便论证这里1美元能产生10倍效益。很多人投身慈善就是想听帮助人的故事。虽然我们也会讲述受益者案例，但更擅长说服分析型的捐赠者——科技行业多这类人。有趣的是，不擅社交的技术人才反而更渴望回馈社会，追求捐赠影响力。

Well, I tend to segment donors into basically heart donors and head donors and Tech for Good doesn't do very well with heart donors. Even if you can make the argument that a dollar in here is gonna have 10 times the impact of a dollar into direct services, there's a lot of people who go into philanthropy because they want to help people and they want to hear about how you're helping people. And so yes, sometimes we can tell the stories of our end user that has benefited from this, but usually we tend to do better with head donors, donors who are more analytical. The tech industry is more full of analytical people. I mean it's great that, you know, tech people who are not always great with people actually see this desire to give back and they want their donations to have a lot of impact.

我们在科技/商业富豪及其基金会方面表现优异，企业捐赠也很成功。'科技向善'通常还有盈利模式——人权领域较难，但教育、医疗总有收入流。很多捐赠者喜欢这种模式：投入几百万，次年二月你就能靠营收维持预算。当前生成式AI热潮下，连原本对技术没兴趣的人都在问'你们的AI战略是什么'——无论好坏。

So we do really well with, you know, people or the foundations of people who've made a lot of money in tech or in business and we do really well with corporations. And tech for good also has the benefit that there's usually a revenue model, I mean human rights, it was harder to come up with a revenue model because there's really no money in human rights, but in you know, but in education, in health, I mean there's usually some kind of revenue stream there, and so there's a lot of donors who like the idea that you know, hey maybe we give you a few million dollars and in February you're making the majority of your budget in revenue. So, but I think that with current Gen AI craze, with people who you wouldn't have thought been interested in funding technology saying, what what's your Gen AI strategy? Mhmm. For better or for worse.

嗯。我们正利用这个契机展示：少量技术投入能为非营利组织带来巨大的社会效益回报。

Mhmm. I we're I think we're taking advantage of a lot of this opening to actually show how a little bit of money into tech for nonprofits can pay really big dividends in terms of social impact.

嗯，那很棒。我想现在非营利顾问的市场也很火热，他们能帮助非营利组织回答这个问题：你们的生成式AI战略是什么？

Mhmm. That's great. Well, and I imagine there's quite a hot market now for nonprofit consultants too who can help the nonprofits answer that question, what is your Gen AI strategy?

确实有，而且你知道，非营利领域里确实有一些优秀的顾问。

There are, and you know, I mean, are some good consultants in the nonprofit sector.

是的，我没有任何贬低的意思。

Yeah, I'm passing no aspersions.

但我认为其中一个挑战在于框架问题，对吧？如果你的框架是如何让你具备生成式AI能力，那你就用错了框架。你等于在说我们有一把大锤子，快找些钉子来敲。这通常效果不好。所以我试着告诉人们：你们有真正的问题吗？

But I think that one of the challenges is like what's the frame, right? If your frame is how are we going to make you Gen AI enabled, you're in the wrong frame, right? You're saying we have this really big hammer, let's go find some nails. It doesn't work very well in general. So I try to tell people like do you have a real problem, right?

比如需要你们服务的人数是你们实际能服务的五倍，这才是问题。现在，我们能想象技术如何帮助解决这个问题吗？生成式AI是最佳技术方案吗？还是有其他更合适的技术？我在书里吐槽过'定制崇拜'现象——那些顾问总说'我们可以为你定制开发'。但年预算百万美元的非营利组织根本不需要花50万开发定制软件，无论他们觉得自己多么独特。

You know, there's five times more people that want your service than you can actually serve. That's a problem. Now, can we imagine how technology can help solve that problem? Is Gen AI the best technology for or is there another technology that's that's the right way to kind of go about that, and so it's the same thing, you know, one of the things that I complain about in the book is the cult of the custom, which is often promoted by consultants who say, oh yeah, we can build that for you. You know, your average million dollar a year nonprofit does not need to spend $500,000 building a custom piece of software, no matter how unique a snowflake they think they are, right?

在这个领域里，你们很可能和其他同行很相似。或许应该彼此合作共建技术平台，因为你们真正的竞争点不在于技术质量，而在于服务质量以及能否筹到资金来开展服务。

I mean, you know, if you're in this field, you're probably a lot like other people who are in this field, you know, maybe you should collaborate with each other and build a common tech platform that will all help solve your problem because you actually don't compete on the quality of your technology. You compete on the quality of your services and, you know, and whether or not you can actually raise the money to do the services.

没错。我认为AI对非营利组织的变革性在于——你书中举了很多例子——不一定要追求光鲜前沿的技术，而是用AI写更好的筹款信、更个性化的感谢函、起草提案书。这些日常运营的琐碎工作若能减轻人员负担，就是实实在在的贡献。

Yeah. Well, I think that's one of the transformative things of AI for nonprofits is, and you give a lot of examples in your book, of it's not necessarily the shiny, new cutting edge piece of technology, but it's using AI to help you write a better fundraising letter, to help you write more custom thank you notes, to help you draft a proposal, you know, just those sort of nuts and bolts pieces of running a nonprofit. If you can take away some of that stress and tedium from the people who have to do that all the time, then that's a real contribution.

是啊，我常开玩笑说我们以前用AI写作就是拼写检查和语法检查。没错，那在当时就是AI。

Yeah. I mean I mean, I I joke that, you know, I mean, the previous thing we used AI for in writing was spell checking and grammar checking. Yes. That was AI at the time. Right?

麦卡锡不是说过吗？一旦技术成熟，我们就不再称它为AI了。

We don't think of it anymore. What is it? McCarthy's quote? You know, once it starts working, we stop calling it AI. You know?

现在这些工具就像是新一代的拼写检查。它们生成的文本能极大加快编辑速度——比如把千字方案压缩成资助申请表要求的250字。当然不能直接粘贴AI结果，但它能提供多种精简思路，可能让你节省一半绞尽脑汁的时间。

But, you know, and so this is kind of like the next generation of spell checking, but, you know, I mean these tools, if you check what they put out, can really help you edit something a lot faster, and of course, you know, one of our favorite applications is, you know, squeeze that thousand word concept note into the two fifty word box you have in the grant application, And you probably shouldn't just paste in the two fifty words, the chat GBT, but you'll probably get a really good idea of several ways how you could squeeze down your concept into those two fifty words, and maybe it takes you half as long as it would if you were trying to do that by just using your brain power.

没错。你在社会正义与技术交叉领域创立过多个项目。对于那些想追求影响力驱动而非纯商业职业的年轻技术人才，你有什么建议？

Yeah, exactly. So you have built ventures at the intersection of social justice and technology. What advice would you give to young technologists who would want to pursue impact driven careers rather than purely commercial ones?

好在大学里这已是学生们的热门话题——这点你应该很清楚。应届毕业生与社会变革前线之间的距离，并不像在商业领域那样遥不可及。比如刚毕业就想去打败思科的路由器业务基本不可能。我的建议是：先掌握所在领域的最佳实践。无论是设计思维、快速原型开发，还是敏捷精益方法——这些硅谷的标杆方法论，即便你将来加入非营利组织也同样适用。对于背负学生贷款的人，我会说先去谷歌这类公司工作还清贷款，然后再以几分之一的薪资投身'科技向善'领域。

Well, the great thing is this is a hot topic in universities among students as I think you know well. And so, and the distance between, I don't know, a fresh grad and the front lines of social change are not as big as they are, you're not likely to go off and beat Cisco at the mega router business like, you know, fresh out of school. And you know, my advice is, you know, learn the best practices in your field, right? If you're into design, human centered design, rapid prototyping, you know, agile, lean, all these things that are now touchstones in Silicon Valley that I would use if I was trying to go through Y Combinator, our big, you know, famous accelerator for for profit ventures, though that skill set is still really valuable in the nonprofit sector. And so I tell people to go do that if they've got big student loans, I say well go work for a Google or someone like that, pay off your loans and then come over to the bright side of technology for us for a fraction of what you were being paid.

关键在于保持这种认知：我们要服务的对象是同样复杂的个体，只是被资本主义或社会机制分配了糟糕的人生剧本。想象一个发展中国家患有残疾的有色人种女性——她的世界充满不公。如果你能思考如何平衡这种不平等，如何打造真正解决她痛点的工具，那就走对路了。我希望人们有机会参与实验室课程设计解决方案，20个创意中总有1个能成长为成功的社会企业。

But I think it's, you know, having that same sophistication about, you know, the people we're trying to serve are sophisticated human beings who happen to have been dealt a, by and large, a raw hand by the way capitalism or society works, right? You know, if you're a woman of color with a disability in the developing world, the world does not look very good for you, right? And if you can think in terms of how can I help equalize this unequal situation, how can I build a tool that is actually helping solve a problem that she actually has, then you're on the right track? And so I hope that, you know, people get a chance to, you know, go through a lab course where they design solutions, you know. One out of 20 of those turns into a successful social venture.

这非常酷。当然不是每个人都适合创业，毕竟整天筹款并不有趣。但我们应该让职业流动正常化——就像律师和医生那样，人们在商业、非营利组织和政府部门之间自由切换。

That's really quite cool. Or you know try to actually get in on the ground floor and not everyone is cut out to be an entrepreneur. I mean, money as your full time job is not the most fun thing to do. But but I think that I we want to normalize the idea of people going from, you know, the for profit world to the nonprofit sector to the government sector and back again. Something that was pretty normal and is pretty normal for us, let's say, attorneys and medical professionals.

科技行业曾偏离这种模式，随着规模膨胀逐渐远离人性。但现在重建这种连接的机会就在眼前：当你亲耳听到因你开发的技术而圆梦大学的学生故事，或是见证人权捍卫者凭借你的工具在国际法庭击败专制政府——这正是'科技向善'工作者坚守的原因，尽管这样的故事远比商业传奇来得稀少。

The tech industry wandered away from that as I think we got further away from people as we built bigger and bigger scale. But I think that that opportunity to engage is really there and feel the direct connection between the technology you're building and the impact it has on society. And that's why people come to work in this sector, the tech for good sectors, because yes, they're applying their craft, their profession, but then they hear the story of the kid who got through college that wouldn't have gotten through college without that piece of technology, or that human rights defender who beat the repressive government in an international, you know, tribunal. Those are the things that we live for because often, you know, those stories are not as common as the stories that go in the other direction.

确实。我们加州大学的学生普遍具有使命驱动特质。虽然也培养大量盯着科技行业的CS专业学生，但当前初级程序员招聘放缓或许能引导部分人才流向社会部门——他们接受的训练本就能让该领域受益。

Yeah. Definitely. Well, I like to think that the students that we have, especially at UC, are quite mission driven. And, of course, we're also cranking out plenty of CS majors and, you know, students who are have an eye on the tech field. But perhaps the slowdown in hiring at the entry levels of computer science might help to redirect some of that talent into the social sector, because I know that that sector would definitely benefit from the training that they that they already have.

我正酝酿一篇专栏：过去几年谁赚得最多？科技巨头。好吧。

I I I this is an op ed that's brewing. Like, you know, who who made more money in the last couple years? The really rich people in tech. Okay. Great.

既然忙着裁员，何不资助更多'科技向善'创新？这些人才自会找到造福社会的途径，世界也将更美好。当然已有少数大捐助者在行动——我们需要更多这样的人。

Why don't you, as while you're busy laying these people off, why don't you actually fund more tech for good innovation and, you know, those people will find their way to doing something good for society and, you know, the world will be a better place. And of course, there are a handful of those big donors who are doing that sort of thing. Yeah. We need more of them.

是的。伯克利的创业文化非常活跃，将这种精神与校长倡导的'变革者'计划结合，或许能培育出更多与你合作过的社会企业。

Yes. Well, and also, I think that combination of tech for good and entrepreneurship, especially at Berkeley, there's a very lively entrepreneurship culture. And trying to marry that with, say, a changemaker kind of program, which is something that our chancellor has really championed, I think is one good path to lead to some of the programs and companies that you have been working with.

没错。我们非营利组织就推荐过员工去哈斯商学院深造——非营利组织的工作经历反而成为他们的优势。我也从伯克利、斯坦福等校招聘过人才。现在商学院里营利与非营利商业计划竞赛的参赛数量已不相上下，至少我希望现状仍是如此。嗯。

Well, yeah. And, of course, you know, I've sent people to Haas, your business school, out of our nonprofit, and I I think that working for a nonprofit positioned them well for graduate I've hired people out of Haas and out of Berkeley and Stanford and Santa Clara and San Jose State because, I mean, I think that, you know, is happening more and more, this sort of evolution. And I I'm often finding out that, you know, the for profit business plan competition and the nonprofit business plan competitions have similar number of entrants in some of these campuses. At least I hope that's still the case. Mhmm.

是的，看到这个很棒。我们已经讨论了很多关于硅谷的话题。那里的精神一直是，并且在某些情况下可能仍然是‘快速行动，打破常规’。那么你对科技造福人类的愿景与这种精神有何不同？

Yeah. That's great to see. So we've talked a lot about Silicon Valley. The the ethos there has been and probably continues to be in some cases, the idea of move fast and break things. So how does your vision of technology for humanity differ from that ethos?

嗯，在非营利领域，我们有道德义务不去打破东西。

Well, think in the nonprofit sector, have an ethical obligation to not break things.

是的。

Yeah.

对吧？你知道，我的意思是，在‘科技向善’领域，我们没有明确的希波克拉底誓言，但我们应该有，对吧？首先不造成伤害，然后确保你实际上在做有益的事情，并且积极管理不良后果，对吧？我认为这是思考的责任。不，我恰好认为这也是在营利世界中打造成功产品的一种方式，但你知道，有一种压力要赚取巨额利润，在某些情况下将现有客户视为通往未来客户的垫脚石。这不是一个非常友好的概念。

Right? You know, I mean, we don't have exactly a Hippocratic oath, in tech for good, we should, right? You know, first do no harm, then actually make sure you're actually, you know, doing things that are well and that you're actually actively managing for the bad outcomes, right? I think that's incumbent upon thinking, no, I happen to think that's a way to build a successful product in the for profit world as well, but you know, there's been this pressure to make immense amounts of money and in some cases to treat your current customers as kind of just a stepping stone to the future customers that you're going to have. It's not a very friendly kind of concept.

因此，我认为‘科技向善’让人们能够做大多数科技人真正想做的事情。我接触过很多对科技行业工作感到失望的人，无论是他们已经在第七家初创公司工作，牺牲了家庭生活却从未成功，因为所有这些初创公司都没有赚到钱，或者即使赚了钱，他们也被骗走了，这在硅谷是一个相当常见的故事。因此，能够从事有意义的工作，过上体面的生活，拥有健康保险，并且让你的孩子能够向朋友解释你在帮助盲人或其他人，这种想法非常有吸引力。我告诉我妻子我将在非营利领域做一年‘科技向善’的事情，然后聘请一位执行董事，我想那是三十五年前的事了。所以，你知道，从事有意义的工作是非常有粘性的。突然之间，你的经济目标从致富变成了谋生，并且过上值得的生活。

And so what I think that Tech for Good allows people to do is actually do what most tech people would want to be doing. I mean I talk to a lot of people who are disillusioned by working in the tech industry, whether it's they're on their seventh startup, they've, you know, sacrificed their family life and they never made it because none of the startups actually made any money, or if it did make money, somehow they got screwed out of it, right, which is quite a common story in Silicon Valley. And so the idea that you could work on something that actually means something, make a decent living, have health insurance, and have your kids be able to explain to their friends that, you know, you help people who are blind or whatever it was. I mean, I told my wife I was gonna do this tech for good thing in the nonprofit sector for a year and then hire an executive director, and I think that was over thirty five years ago. So, so you know, it's very sticky to be working on something that means something And, you know, and suddenly your economic objectives shift from getting rich to making a living and and having a life that's, well, worth worth worth living, I think.

是的，并且产生影响。我认为你参与的许多项目确实产生了影响，而且它们是棘手的问题。所以对某种人来说，这也很有吸引力，对吧？

Yeah. And having an impact. I think a lot of the programs that you have worked on have certainly had an impact and they're they're wicked problems. So for a certain kind of person, that's also quite attractive. Right?

在我们的有生之年，你无法解决贫困问题。所以总会有其他事情可以努力去改变。

You're not going to solve poverty in our lifetime. So there's always gonna be something else to to work on to try to make a difference.

嗯，这种迭代的方法是解决棘手问题的好方式，对吧，因为棘手问题之所以复杂是因为它们不直接，对吧，所以如果你可以尝试一个项目，因为我们的技术总是在提供一种程序化的干预。必须有人想过这个人需要什么，而我们使用技术作为实现这一目标的工具，但如果你认为这个东西真的能帮助人们，一年后你意识到60%的人完全没有得到帮助，而40%的人得到了帮助，你会变得更聪明，好吧，我们需要确保我们只关注我们能帮助的人，而不是那些我们无法帮助的人，或者嘿，我们有一个新想法也许能帮助那60%我们错过的人。拥有这种迭代的方法，你知道，这是商业现在对技术采取的基本方法，在社会领域也同样适用。只是我们优化的不是每单位努力的最大金钱收益，而是在某种程度上的收支平衡的同时产生最大的影响。

Well, and this iterative approach is is is a good way to tackle wicked problems, right, because wicked problems are complicated because they're not straightforward, right, and so so if you can try out a program, because I mean our technology is always, you know, basically delivering a programmatic intervention. Someone has to have thought of what is the thing that this person needs and we're using technology as a vehicle to do that, but if it turns out that you think this thing is going to really help people and you get a year in and you realize that, you know, 60% of people aren't helped at all and 40% are, you'll start getting smarter about, okay, we need to make sure that we only focus on the people we can help and less on the people we can't help, or hey we have a new idea that can maybe help those 60% we missed out on. Having that kind of iterative approach, is you know the fundamental approach that business is now taking to technology, is great in the social sector as well. It's just that our, what we're optimizing for is not extracting maximum dollar per unit of effort, but instead having the maximum impact while somehow breaking even.

是的。完全正确。这就是非营利组织的成功。

Yeah. Exactly. That's success for the nonprofit.

我们会接受的。绝对。

We'll take it. Absolutely.

那么，能否请您分享一个工作中的故事，最能体现技术以主流科技叙事中不常见的方式改变生活的力量？另外，您书中提到的‘数据更好交易’项目，或许可以谈谈您正在推进的数据隐私管理工作。

So could you share a story from your work that best illustrates the power of technology to transform lives in ways that are not usually seen in mainstream tech narratives? And one example that I just wanted to call out from your book too, maybe you could speak to, is a project on a better deal for data. So talk about the data privacy management that you've been working on.

好的，我认为关于残障人士的无障碍技术，有个非常典型的例子——在我刚入行时，盲人主要通过他人朗读或磁带有声书获取内容（没错，就是实体录音带）。而在我职业生涯中，我们实现了从磁带书到自主扫描书籍的跨越。我记得有位新闻系研究生曾告诉我，她原本打算退课，因为需要批改同学作业却总找不到朗读者及时完成任务，直到有了扫描朗读设备配合语音电脑才解决了问题。今天董事会还听到类似案例：阅读障碍儿童从厌恶阅读到爱上阅读的转变。

Yeah, I think, I mean, one of the things about accessibility for people with disabilities, it's very sticky is we've moved from a world where when I started by and large blind people were read to, you know, often in person or by getting a book on tape and literally tape, right, audio cassette tape. And if you think about during my career in this field, we've gone from books on tape to scanning your own books and you know, like within the first year I remember hearing from a journalism grad student who said I was about to drop out of my journalism class because I had to read my classmates assignments and grade them as part of and I could never get a reader who could actually do it in time to make the deadline, and now that I have a scanning reading machine I can actually scan it and do the work with you know my talking PC. And so you know just that one idea of here and I mean and I got a story today in a board meeting that was very similar to you know dyslexic kids who hate reading suddenly finding themselves liking reading.

想想看，从憎恶阅读到享受阅读，学校自然会变得更友好，学业进步也更可能实现。回到主题——我们从依赖他人朗读、邮寄磁带，发展到自主扫描，再到受Napster启发的Bookshare（虽然当初有人反对这个名字）。现在盲人社群正在共建图书馆：由盲人、阅读障碍者或教师决定藏书，而非视力正常者。当前我们正推动出版业将商业电子书做得像我们改造的书籍一样无障碍，越来越多出版商已在行动。

Well somehow if you go from hating reading to liking reading, something tells me that school is gonna be a much more friendly place for you and you're much more likely to advance in school. But back to the sort of narrative, we went to, you know, from people being read to or tapes being sent through the mail to people doing their own scanning to Bookshare which was mauled after Napster because the woman who was ordering Napster lived two doors down and I got infected by Napster and I went, oh Bookshare, and so that's where Bookshare, I was told not to call it Bookshare, became, oh, the blind community is building a library for each other. Right? If a lot now a sighted person doesn't decide whether it deserves to be in the library for the blind, a blind person decides or a dyslexic person decides or the teacher of someone decides, and now we're in the think the final stages which we've gone to the publishing industry and say, hey you guys are into e books now, why don't you make your commercial e book just as accessible as we've been retrofitting your books? And now more and more publishers are doing that.

如今标准电子书已能满足残障人士需求，他们不再需要特殊版本，主流版本就能完美适配。何况很多人本来就喜欢有声书。

So now the idea is that the standard book that more and more people are getting works perfectly well for people with disabilities. They don't need a special version of it anymore because the mainstream version works. And a lot of people like audiobooks.

确实，我可以作证。

Yes. I can attest.

说到这个...对了，还有‘数据更好交易’项目。好的。

So that that kind oh, and the better deal for data. Yes. Okay.

在结束这个话题前，我想延伸一下：未来当我们能定制合成语音时——比如选择让汤姆·汉克斯为你朗读...

So Before we move on from this example, I just would play it forward a little bit, and in the days when we will soon have synthetic voices, like you could just dial up to have Tom Hanks read you.

这技术现在已经实现了！但愿汤姆·汉克斯能收到版税。其实最近就有类似斯嘉丽·约翰逊的声音争议...虽然不完全是她本人。

Oh, that works today. Hopefully, Tom Hanks got a royalty. Exactly. Kind mean, Scarlett of that problem going on a little bit. Well, it's not exactly Scarlett Johansson.

但听起来简直和斯嘉丽·约翰逊一模一样。试想如果某人失去了嗓音呢？

Just sounds just like Scarlett Johansson. Yeah. So but, you know, that ability, what if you've lost your voice?

确实如此。

That too.

你知道，你还有自己曾经能说话时的录音。

You know, and you had recordings of when you had a voice.

是的。

Yes.

现在你可以恢复自己的说话声音了。

And now you can get back your speaking voice.

没错。太棒了。重大突破，对吧？

Yeah. Amazing. Big. Right?

是啊。所以说这就是无障碍科技领域发生的许多激动人心的事。要知道，在我刚入行时，人们愿意花5万美元买一台Kurzweil阅读机，后来我们的突破是将价格降到5000美元。现在这功能每部智能手机都内置了，对吧？一旦成本降到这么低，那些原本只有严重残障人士——比如无法说话或失明者——才迫切需要的功能，突然就变成了我们所有人都视为理所当然的超级便利功能，比如对着Siri说话或让Google给我的咖啡或茶计时什么的。

Yeah. And and so and this is this is a lot of the exciting stuff that goes on in the accessibility world. You know, people were willing to spend, you know, $50,000 on a Kurzweil reading machine when I was starting, and then our breakthrough was to bring it down to $5,000. Now it's built into every smartphone, right? And once it gets that cheap, suddenly all these things that were desperately needed by someone who has a significant disability, can't speak or can't see or whatever it might be, now becomes this super convenient feature that we all just take for granted that of course I talk to Siri or Google times my coffee or my tea or whatever, know.

那些原本为战斗机飞行员或残障人士开发的技术，正是语音识别的起源，对吧？所以我特别期待看到更多这样的创新进入主流，改善每个人的生活。因为说实话，所谓的'普通人'根本不存在。我们的能力差异巨大，比如开车时的行为能力就完全不同于窝在安乐椅里的时候。但我们都希望能获得更好的解决方案

That's, that those stuff, that stuff that was either built for fighter pilots or people with disabilities was where voice recognition got its start, right? And so, so yeah, so I'm really excited about this moving more and more of these innovations into the mainstream to make everyone's lives better because frankly the average human being, you know, is not really a real thing. We have a gigantic range of capabilities and often you know when you're driving you're like you have a different set of disabilities than when you're seated in an easy chair. But you want to move on to the better deal

是的，请继续。

Yes, go ahead.

好的。这个议题以两种不同方式浮现——通常当同一个想法多次出现时，我就会觉得该重视了。第一次是我们刚开始涉足气候NAG领域时，首场用户访谈是在非洲地方领袖会议上进行的。参会者基本都是非洲各县区级的地方领导人。我们采访的第一位对象叫Kamal Nbogo，他相当于当地'守护太浩湖'组织的负责人——就像我们这边保护太浩湖的机构。他守护的是奈瓦沙湖，那是东非大裂谷区的淡水湖，算是他们的'太浩湖'。

Okay. So this came up two different ways and often it's when the same idea appears multiple times I go, oh, maybe we should pay attention. So the first time when we started working in the climate NAG space, our first interview with users was going to an African local leader conference. So this is almost all people in Africa who are local leaders at sort of the county or district kind of level and we're interviewing first people and the first interview I did was with a man named Kamal Nbogo who led basically the equivalent of Keep Tahoe Blue, which is you know preserve Lake Tahoe here. He did it for Lake Naivasha which is kind of the freshwater lake in the Rift Valley that's kind of their Lake Tahoe.

他当时说：'很高兴你们从硅谷来帮忙，但我们有些问题。'我问什么问题？他说：'我们拥有裂谷区最稳定的生态系统，但我和其他地方领导人都无法获取相关数据。Jim，这叫数据殖民主义。'我当时有点懵，什么？

And he basically said, hey you know I'm really glad you're here from Silicon Valley, you're offering to help me, but we have some problems. I'm like, What? Well, we're the best steady ecosystem in the Rift Valley and I and all the other local leaders don't have any access to that data. He said, Jim, that's called data colonialism. And I was a little bit, what?

他说：'等等，你从硅谷来却不知道数据殖民主义这个词？'呃...确实不知道。虽然他措辞比这委婉得多，但我明显感受到了连环暴击般的信号：赶紧去补课吧。

He's like, wait a minute. You're from Silicon Valley and you don't know the term data colonialism? Mhmm. No. He was much more polite than that, but I was getting the whack, whack, whack kind of signal, like go study it up.

结果发现确实存在这样一个学术研究课题。我当时就想，天哪，我们在数据使用上简直带着殖民主义思维。大约一年后，我和美国土壤研究的主要资助者——食品与农业研究基金会的Lakisha Odom博士交谈时，她对我和项目负责人Steve Francis说：'你们需要为农民创立类似HIPAA的法案'。

It turned out that there was such a, you know, was an academic thing that had been studied. I'm like, oh my God, we're being kind of colonial about how we use data. And then like a year later, I was talking to one of the major funders of soil research in The United States, Doctor. Lakisha Odom from the Foundation for Food and Agriculture Research, and she said to me and to Steve Francis who runs that program, You guys need to invent HIPAA for farmers. Okay.

我第一反应是：HIPAA？那多麻烦啊。但她解释道：'不，你明白吗？就像你去看医生时讨论隐私问题，你获得帮助的同时，也清楚你的数据可能会被用于医学研究，为解决你、你的孩子和类似人群的问题做贡献——这种社会契约大家是能接受的。'

And I went, HIPAA? That's kind of a pain. And she's like, No, no, but you get it, right? You go in and you talk to your doctor about something that's confidential. You get helped, and you know that your data is gonna be probably used in medical research to solve that problem for you and people like you and your kids, and the social bargain, people are okay with it.

她指出当前农民抵触数据收集，这严重阻碍了农业研究的未来发展。根本原因在于，农民从经验中得知数据被采集后往往会吃亏。虽然她措辞委婉，但意思很明确。这促使我们提出'数据公平交易'理念——正如Shoshana Zuboff教授所说的'监控资本主义'导致人们认为数据采集就意味着被剥削，对吧？人们发现数据总被用来从他们身上榨取更多利益，而他们的权益却被忽视。这类现象比比皆是。

She says right now farmers are not wanting to cooperate with data collection, which is killing the vision of the future of ag research, and she said the reason is that when people collect data from them, they've learned that by and large they get screwed. She again did not say that, but she said something that was its equivalent. And so that's where the better deal for data came from, which is, you know, surveillance capitalism as it was deemed by Professor Sarzana Zuboff, is leading to these things where people are getting the idea that when data is collected from them, they're getting abused, right? People are figuring out a way to make more money off of them and their best interests are not paramount. And we've seen this a lot, right?

所以我们思考：不想出卖用户的非营利机构和企业该如何应对？我们多方寻找模板却一无所获。最终方案是制定几条简单承诺：不出售数据给中介、数据使用以用户利益为先、用户掌握主导权。若进行数据研究，会做去标识化处理并要求不得重新识别——这很关键。本质上是建立信任契约：我们采集数据只为帮助你及同类人群、推动科研，而非靠你的敏感数据牟利。毕竟非营利机构收集的往往是弱势群体最脆弱的信息——比如遭受性侵儿童的数据。

So the idea is how can nonprofits and businesses that don't want to sell out their users, how do they do that? And we went around and we looked for a model and there wasn't one. So our idea was what if you could come up with a handful of simple promises that says, you know, we won't sell your data to the data brokers, we're going to use the data primarily to benefit you, you're in charge, you know, if we do any research on your data we're gonna de identify it and get people to promise not to re identify it, which can be a problem, and so the idea is like we're gonna make this bargain that you can trust us that if we're gonna collect this data it's because we're helping you, helping people like you, doing science, but not trying to become wealthy on the backs of your data, especially your sensitive data, because I mean the nonprofit sector gathers data from vulnerable people about the things that make them vulnerable. It's in a lot of ways the most sensitive data you can imagine. Data on kids that have been sexually assaulted.

我们绝不能从性侵受害者身上牟利——而这正是当下硅谷大规模进行的勾当，他们一边借此盈利一边假装防治。我们这个领域的工作者应该承诺杜绝此类恶行。非营利机构领导者都渴望能签署这样的可信承诺。与忧心忡忡的农民沟通时，我们希望——当然还需验证——只要承诺不把数据出卖给农业巨头，他们就愿意配合研究。

We need to be not making money off of sexual assault, which Silicon Valley today does at scale, make money off of that and works at reducing the amount of it. So I think those of us who work in these areas can commit to not doing any of that kind of bad stuff and when I talk to nonprofit leaders, they would love to sign up for something they could say, yes, we'll be trustworthy. And we talked to farmers who are very, very worried about this. We're hoping, we'll see, that they'll be willing to cooperate with ag research as long as we don't sell them out to big agribusiness.

这个模式很有意思。大学里机构审查委员会的问卷也会涉及类似问题：确保数据处理的合规性、采集知情权及后续使用规范。我认为特定领域的非营利组织确实需要标准化模型来声明'这是我们处理数据的方式'。

Yeah. No, that's a really interesting model. Some of those questions come up if you're in a university setting in an institutional review board kind of questionnaire to make sure that you're treating data appropriately, that you have consent for gathering data and how the data is gonna be treated after you collect it, that I think it makes a lot of sense to have some kind of standardized model also for the nonprofit organizations in a particular sector to say this is how we are going to treat your data.

没错。高校人类受试研究有严格的伦理要求，我们不是要全盘照搬到非营利领域，而是建立介于两者之间的基准线。这样既能维持现有工作，又能承诺数据保护与使用限制。若涉及研究，必须执行知情同意——数据越敏感，伦理要求就该越高。

Yeah. I mean I mean, doing human subject research in universities has very high ethical requirements, and our idea is probably not to replicate those in the nonprofit sector, but it'll build a floor that's maybe halfway between those or something like that so that they can continue doing what they're doing, but make some promises to safeguard that data and what not to do with it. And if you are doing research you should also do informed consent and all the things that, the more sensitive the data, the more likely you are to have a higher ethical requirement on the data you're using.

明白了，非常感谢。

Yeah. Great. Thank you for that.

应该的。

Sure.

慈善机构、政府和私营部门都在影响技术发展。您认为当前谁最需要加大作为？

So philanthropy, government, and the private sector each play a role in shaping technology. In your view, who needs to step up the most right now?

慈善事业。我是说，我想说政府可以做得更多，但现在似乎不是政府监管的时候，而且资金明显大幅减少。所以我对慈善事业抱有希望。我认为有少数人已经站出来了。总的来说，我认为美国的大部分慈善机构都保持低调。

Philanthropy. I mean I mean, I would like to say that government could do more, but this does not seem to be a time for government regulation and obviously a lot less funding. So I'm hopeful about philanthropy. I think a few people have stepped forward. I think by and large, most of say American philanthropy is lying low.

实际上，在与同行交流中，我们发现过去一年私人慈善有所收缩，这很可怕，因为你知道，那么多公共资金被搁置了。不仅仅是外援资金——显然这上了很多头条——还有很多国内资金也被搁置，许多州政府及地方政府暂停了资金计划，因为他们不确定预算会因为州或联邦层面的经济放缓而被削减多少。所以一切都像是冻结了，所以我希望慈善事业——虽然它不可能填补已出现的缺口——但至少能资助很多这类事情的重建。尽管，你知道，这很不幸，但据我所知，很少有事情能让你用少得多的钱做更多事。科技恰好是其中之一，尤其是当你主要用它来创造社会效益时。

Actually talking to my peers, we've seen a contraction in private philanthropy in the last year, which is scary given that, you know, so much of public funding is on hold. And it's not just foreign aid funding, obviously that's gotten a lot of the headlines, but a lot of domestic funding has been put on hold, a lot of state and government and local government are putting funding plans on hold because they're not sure how much their budget's gonna be cut because of slowdowns from state or federal levels. So everything is kind of freezing up, and so I'm hoping that philanthropy, which cannot possibly fill the gap that's been created and, but is there to kind of, I don't know, fund the reconstruction of a lot of these things. Though, you know, it's unfortunate, but there are very few things I know that allow you to do more with a lot less money. Technology happens to be one of them especially if you're doing it primarily to do social benefit.

是的，谢谢。我很感激。

Yeah, thank you. I appreciate that.

但我看到我们的主持人有点在周围转悠。

But I see our host circling around a little bit.

是的。所以如果有问题，请写下来。我们这里还有一个问题，然后我们会请杰拉尔德上来提出一些观众的问题。好的。那么展望未来，是什么让你最乐观地认为科技能够应对我们最大的全球挑战？

Yeah. So if you have questions, please write them down. We'll have one more question here, and then we'll ask Gerald to, come up and offer some questions from the audience. Great. So looking ahead, what gives you the greatest optimism that technology can rise to meet our biggest global challenges?

可能是气候变化、不平等、虚假信息。那么，什么给了你最大的希望，又是什么让你夜不能寐？

That might be climate change, inequality, disinformation. So what gives you the greatest hope and what keeps you up at night?

嗯，我是说，你知道，我仍然是个技术乐观主义者，尽管我花了很多时间劝人们不要以愚蠢的方式使用技术，对吧？这有点像我的角色。但我是说，我对技术能做什么感到兴奋。我的很多工作，你知道，还有数百人在其他组织也做这类工作，我们所做的就是身处科技领域，了解最新创新，看着企业家和投资者挥霍数万亿美元，然后找出哪些是有意义的，同时我们也在观察社会领域未满足的需求，并尝试将这两者结合起来。对我来说，最有趣的就是尝试去连接这两者，无论是告诉人们他们应该和谁谈谈——谁在做类似的事情，还是实际上，你知道，500次谈话中有一次可能就是我的下一个社会企业。

Well, I mean, I am, you know, still a technical optimist, even though I spend a lot of my time talking people out of using technology in dumb ways, right? That's kind of my role. But I mean, I get excited about what technology can do. I mean, a lot of what my role is, and again, you know, I talk about hundreds of people who do this kind of work in other organizations as well, what we get to do is we are all in the tech field hearing about the latest innovations, watching entrepreneurs and investors blow trillions of dollars and figuring out what makes sense, and then we are watching that and we're watching what's going on in the social sector, unmet needs, and get to put those things together. For me that's like the most fun thing I get to do is try to go, I can connect these two things and whether that's tell people who they should be talking to who's working on something like that or actually, you know, one out of 500 conversations is my next social enterprise.

不过我觉得我得停下来了。我得开始放慢脚步，让下一代接手。这就是为什么我希望人们读这本书。让我犹豫的是什么？不受约束的资本主义往往会伤害很多人，而且，你知道，我认为我们作为一个社会需要持续施压，无论是通过监管——这在一段时间内会是个问题领域——还是仅仅通过社会压力。

Though I think I've got to stop. I got to start slowing down and get the next generation going. That's why I want to get people to read the book. What gives me pause? Unfettered capitalism tends to step on a lot of people and and, you know, I think we as a society need to keep putting pressure whether that's through regulation, which is gonna be a problematic area for a while, or just social pressure, you know.

你们对孩子们做了可怕的事情，我们恨你们，这些实际上会让公司改变他们的行为，我觉得在孩子们身上，但你知道，我是说，你知道，法国的巴黎银行基本上因为难堪而退出了对石油和天然气项目的贷款。现在社会能维持这种压力吗？我认为社会可以继续对关心的事情施加压力，无论是通过民选官员和监管，还是仅仅通过社会压力，我希望他们对那些变得极其富有的人施加更多压力——你知道，这往往是以一定的社会成本为代价的，而我认为这些成本本应被最小化。

You guys are doing terrible things to kids and we hate you, and these things actually get companies to change their behaviors and I feel in kids, but you know, I mean you know, BNP in France got basically embarrassed into getting out of lending to oil and gas projects. Now can society sustain that? I think society can continue to put pressure with things that society cares about, whether that's through elected officials and regulation or through just putting social pressure, and I hope that they put more pressure on people who become, you know, immensely wealthy often at a, you know, some social cost which was not minimized to the extent I thought they should have.

是的，谢谢。嗯，他们也可以利用技术工具来做这件事，比如赋能倡导组织，组织社区工具。

Yeah, thank you. Well and they can also use technology tools to do that so enabling advocacy organizations, organizing community tools.

好吧，我...我会...我不会。既然我刚像拆解了硅谷的大部分神话，我想强调的是，当我走进一家科技公司说'我想要免费授权使用你们的核心技术知识产权，去开拓你们不关心的市场——我可以把这技术带到赞比亚，带给残疾儿童，带给人权活动家'时，80%的情况下他们会同意。人们总是很惊讶：那些贪婪的家伙怎么会？但事实就是如此。

Well, and and I'll I'll do I'll no. Now that I've just, like, torn down most of most of Silicon Valley, something I I I wanna emphasize is when I go to a tech company and I say, I want a free license to your crown jewels, your intellectual property, so I can go after this market you don't care about. I can bring this technology to Zambia, I can bring it to disabled kids, I can bring it to human rights activists. They say yes, 80% of the time. And people are surprised like what those greedy so and so's, and I'm like, no, no.

他们真心为自己创造的东西感到骄傲，却因现行体制阻碍他们做有益社会的事而沮丧。所以我经常告诉学生们——特别是当我给班级讲课时——要知道大多数毕业生会进入商界。当你掌管某项知识产权时，如果有人提出'我们可以授权把它带到你们根本不关心的市场'，你应该答应。我相信很多人内心都有行善的愿望，有为自身创造感到自豪的本能。我们只需创造更多渠道，让他们将成果惠及被现有资本主义模式忽视的那90%人类。

They're really proud of what they've created and find it frustrating that our system makes it very hard for them to do a socially beneficial thing. And so a lot of ways, you know, when I tell, especially when I go to talk to classes, right, I know that most of the class, the average class is going to go into business. Like when you're in charge of some intellectual property and someone says, hey, we could license that and take it to this market that you don't even care about, you should say yes. And so I think that in the heart of a lot of people, the desire to do good is their pride in what they've done is there. We just need to create more ways for them to channel what they've created into benefiting the 90% of humanity that is neglected by our existing capitalist model.

谢谢。这是个很好的暂停点，我们请杰拉尔德看看观众是否有问题要提问。

Thank you. It's a good note to pause on and we'll ask Gerald if you have some questions from our audience.

太好了，感谢观众提出这些精彩问题，我想请两位共同探讨。先从第一个问题开始：技术跨越确实存在，你们认为数字公共基础设施的兴起是炒作还是现实？对此有何担忧？

Great, well thank you audience for some wonderful questions and I wanna direct them to both of you guys to kinda kick around. Let me just start with this first one here. The question is technological leapfrogging is real. Do you believe that the rise of digital public infrastructure, Is it hype or real any concerns?

你想回答这个问题吗？

You want to grab that one?

我可以先开始。公共利益技术这个领域确实值得探讨，吉姆刚才提到的很多内容，在福特基金会等资助方的支持下，已经在投资开发和部署解决社会问题的技术方面取得进展，这部分回答了这个问题。

I can start. Well, field of public interest technology is one that we could address and this kind of phrasing and a lot of what Jim has already talked about has come about and with some funding say from the Ford Foundation and other funders, has really done some of this investing in the development and deployment of technology that's going to address some of the social sector issues. That goes partway to answering that question.

没错。要知道福特基金会是在模仿五十年前创建公益法律运动的模式。当然这本书讲的就是公共利益技术——虽然我用的是'科技向善'的表述而非公共利益技术。

Yeah. And and and and, of course, you know, think Ford is modeling that after helping create the public law movement, whatever it was fifty years ago. And and, of course, this is a public interest technology book. Right? You know, mean, even though I use the tech for good framing as opposed to public interest technology.

我认为数字公共基础设施运动本质上是推动开源基础设施，尤其在国际层面。联合国一直在强力推进，越来越多国家政府表示'我们要掌握自己的技术未来'。这在公共卫生领域已取得显著成效，对吧？

I think that the digital public infrastructure movement is really a push for open source infrastructure, especially in the international context. So the UN has been pushing it pretty hard and increasingly national governments are saying we want to own our own technology future. And this has really been successful in public health. Mhmm. Right?

现在发展中国家的很多技术基础设施都建立在开源软件基础上，比如适配尼日利亚需求的改造工作主要由尼日利亚人完成——不知为何这很受该国政客欢迎。我不认为这是炒作，但它比人们想象的更难：如果完全采用开源技术栈，你就无法像硅谷初创公司那样利用亚马逊AWS或微软Azure等现成基础设施，必须下沉几层基于更基础的组件开发。举个例子...

A lot of the tech infrastructure now in developing countries likely built on top of open source software where the jobs of adapting the software to the needs of say Nigeria is the jobs are mainly filled by Nigerians, and for some reason that's popular with Nigerian politicians. And so I wouldn't say it's hype, it's harder to do than people think, and so of this is that if you're going to be a completely open source stack, you have to give up on benefiting from some of the infrastructure that your average startup in Silicon Valley would be exploiting, right? They'd be building on top of Amazon Web Services or Microsoft Azure. They'd be using all the latest tools. If you're going to build on top of open source, you have to kind of go down a couple layers and build on more basic things, but I'll give you an example.

我们有个完全开源的项目在发展中国家非常流行，比如在印度应用广泛且毫无政治争议。但另一个项目在肯尼亚推广受阻，因为肯尼亚政府要求所有关于肯尼亚儿童的数据必须留在境内，不得触及美国。这确实是个现实痛点——尽管对该项目不利。如果有听众愿意资助一百万美元开发开源版本解决这个问题就太好了。

We have a project that is completely open source, very popular in the developing world, and you know getting a ton of use in India say, and there's no political issues about it because it's open source. We have another project where we're struggling for Kenya to adopt it because the Kenyan government wants to have all the data about Kenyan kids in this case to be in Kenya and not touch The United States. And so I'm seeing it as a very real pressure point, and even though it disadvantages that particular project, I would love to find a million dollars to make an open source version of that and get around that problem just in case anyone in the audience wants to help me with that.

嘿，这里是旧金山。

Hey. This is San Francisco.

我很高兴你提到了开放问题，因为我认为开放数据运动在多年前确实很有帮助，那时你可以获取开放数据用于公民改善项目、健康项目以及提升教育成果。在此基础上，又发展出了开源软件和其他类型的产品。

Well, I'm glad you did touch on the on the open question because I think the open data movement was really helpful a number of years ago when you could get access to open data for civic improvement projects, for health projects, for improving education outcomes. And then just building on that, you have open source software and other kinds of products that kind of evolved out of that.

是啊。

Yeah.

很好。接下来是第二个很棒的问题。关于小型语言模型的看法。他提到了SLM，我想这就是你的意思。

Great. Okay. Second really good question here. Thoughts on small language models. He has SLM, but I think that's what you mean by that.

最少需要多少token才能使机器更好地理解不同的影响领域？这在旧金山是个热门话题。OpenAI和Anthropic不考虑少于10亿token的数据集，这大约相当于150万美元。那么你对小型语言模型有什么看法？

How many tokens at minimum to enable machines to better understand different impact verticals. That's pretty this is San Francisco. OpenAI Anthropic doesn't look into datasets less than 1,000,000,000, which is approximately 1 and a half million bucks. So what's your thought on the small language model?

我不会给出具体的token数字，但可以告诉你我们的做法。我们受益于科技公司的免费技术。目前我们使用OpenAI进行摘要生成，并与其协商了零数据保留政策，因为这是医疗数据，他们对医疗数据有特殊规定。所以我们对此感到满意，当然他们是免费提供的。

So I'm not going to name a token number for some, you know, but but I think that I'll tell you what we're doing, right? We benefit from free technology from the tech companies. Right now we have, we use OpenAI's to do summarization. We've also negotiated a zero data retention policy with them because it's healthcare data and they have special provisions for healthcare data. So we feel good about that and of course they're giving it to us for free.

我认为在北美这样的地方使用这类商业产品的成本可能是可行的，但在非洲这样的地方就不切实际了，太昂贵了。当人们提到小型语言模型时，他们指的是那些最新的大型语言模型（如ChatGPT）在九个月前的水准，那时你可以在个人电脑上运行它们。我认为很多社会公益部门会放弃处于生成式AI最前沿、昂贵且耗能的状态，转而采用一些足够好用、价格合理的小型模型，并掌握数据控制权，因为很多非营利组织宁愿完全不分享他们的数据。

I think that the cost of using a commercial product like that is probably practical in a place like North America, it's impractical in a place like Africa, it's just too expensive. And so, but and when someone says small language model, what they mean is that, you know, the latest large language models, you know, ChatGPT, you know, in the arms race with Claude and everybody else, you know, if you go back to where they were nine months ago, you can get that to run on your PC, right? And so I think that a lot of the social good sector is going to trade being at the absolute bleeding, expensive, energy intensive end of generative AI and trade it off for some smaller models that work well enough and are affordable and give them control over the data because I think a lot of nonprofits would rather just simply not share their data. We certainly have helplines for example that just say, no, don't care if you negotiated zero retention with them. I don't want my data going to them at all ever.

这是一个我们必须支持的原则立场。

And it's a principal position that we have to I think also support.

我想这正好引出了下一个问题：有人问数据的再利用是否包括忽略文化意义的数据解读？

I think this leads right into the very next question here, which is someone says reuses of data, that that does that include interpretation of data omitting the cultural significance?

数据的再利用。

Reuses of data.

是的。这包括当你省略某些标识符或问题时对数据的解释吗？我认为这正是这个问题的本质所在。

Yeah. Does this include the interpretation of the data when you are omitting some of the, you know, identifiers or issues? I think that's the nature of this question, yeah.

我来开个头。谢谢。因为我觉得这触及了核心问题。数据殖民主义的特征之一就是将数据主体和社区与数据使用者和分析者割裂开来。这往往意味着脱离具体语境来处理数据。

I'll get it started. Thank you. Because I think it's getting in the heart of something. So one of the characteristics of data colonialism is separating the data subjects and the communities from who's using the data and who's analyzing it. And often that means removing it from the context.

举个简单例子——我要谈谈慈善领域的按绩效付费。我们说：如果你给很多孩子接种疫苗，我们就支付这么多钱。结果呢？有些地区报告显示可接种儿童覆盖率100%，但实地调查发现实际接种率只有60%。为什么？

And a simplistic example is, and this is, you know, I'm going talk about philanthropy, pay for performance. We want to pay you this much money if you vaccinate a lot of kids. What happens? Well some districts have one hundred percent and of the kids that could possibly be vaccinated according to the reports, but then if you go survey the village sixty percent of the kids are vaccinated. Why is that?

因为你创造了数据造假的动机。另一个例子：看这个发展中国家各省数据——某省达成率只有目标的一半，我们要削减拨款。但如果背景是该省刚遭遇台风袭击，半数医疗设施被毁呢？脱离背景分析数据做决策，却对实际情况一无所知，这就是不公正的殖民主义行为。

Because you created an incentive to lie about the data, so that's one thing. Another another example of this is, hey look, look at all the different provinces around this developing country. Oh, that province, they had half of the results that we desired, we're gonna cut their funding. Well, what if the context was that was the province that was hit by the typhoon and half of their medical facilities were destroyed? If you separate the analysis of the data and you come up with your decision, but you don't have a clue about what was going on there, that's unjust and that is a very colonial behavior that's going on.

因此'如何负责任地使用数据'是个关键议题。我们在'更优数据协议'中的工作，正是基于过去十年学术界关于数据治理的研究成果。

So this, how do we use data responsibly is I think a link in and I know this is something, I mean I've learned a lot from the academic field about sort of much of what we're doing in terms of Better Deal for Data is building on all the academic work about data governance that's been going on in the last ten years.

对。这让我想到另一个相关案例——人们使用错误代理指标做决策。最近有个案例：医保服务和保险供给决策是基于某人群实际使用医疗服务的数量，而非其真实需求。他们看到黑人群体医疗服务使用率低，就认为他们不需要更多服务。而实际上低使用率是因为无法预约或无力支付共付额等障碍，算法却无法理解这种复杂性。

Yeah. A related example that calls to mind is when people are using false proxies for for making decisions. So there was a case recently where some health care decisions and health care insurance and provision decisions were being made based on how much health care a certain population was using versus how much health care they actually needed. And so they were looking at, at black people, in this case, thinking that they didn't actually need more health care because they weren't using it. When, of course, the reason they weren't using it wasn't because they didn't need it, was because they couldn't get to an appointment or they couldn't afford the copay or there could be any number of other reasons why they weren't taking advantage of it, but the the algorithm wasn't understanding it in that kind of way.

这导致了差异性影响。

And so that led to disparate impacts.

没错。这正是许多行业AI产品的通病。七年前我们研究人力资源技术时，所有供应商都说：我们的AI工具不歧视女性、不歧视有色人种、不歧视移民或LGBT群体。但当你问'你们测试过这些情况吗'，他们回答'不，我们不测量那些'——你们难道没听说过代理变量吗？

Yeah. That's the story of a lot of AI driven products in industry. We we did a study of human resources technology seven years ago. All the vendors and, of course, all of human resources is filled with AI tools. We don't discriminate against women, we don't discriminate against brown people, we don't discriminate against, you know, immigrants or LGBT people and you say, oh great, have you tested for those things?

早期典型案例是亚马逊开发的面试筛选工具，由于训练数据主要来自男性员工，导致对女性的歧视严重到无法修复。

And they're like, well no, we don't we don't measure that. Like you have you ever heard of proxies? You know? I mean, I mean, I remember an early example is Amazon tried to train a tool for deciding who to get an interview at Amazon, and the thing discriminated against women so radically it was unfixable because why? Because they were training on a mostly male workforce.

那些已经为他们工作的人。

Who already worked at them.

你知道，我是说，我们可以——举个例子，我非营利组织下属的一个小组后来独立成了人权数据分析组，他们指出奥克兰的预测性警务存在这样的问题：毒品犯罪在富裕社区和贫困社区都有发生，但警方前往贫困社区的频率却是皮埃蒙特这类地区的两三倍。这是犯罪率差异导致的吗？不，这是因为AI训练数据本身就基于——当然啦——我们本就存在的种族歧视性警务行为。

You know, and I mean, we could, I mean, another example from one of the groups that was at one of my nonprofits and spun out the Human Right Data Analysis Group, they showed that predictive policing in Oakland, you know, drug crime exists in the population in the wealthy part of town and the poor part of town, but the police go to the poor part of town two or three times as much as they go to Piedmont or places like that. And what, is that because of differential in crime? No, it's because the AI was trained on, well, of course we have racist policing.

没错。原本逮捕率最高的地方是哪里？

Yeah. Where were the most arrests already?

对。所以这同样解释了为什么你必须审阅生成式AI的输出内容。

Yeah. So, I mean, this is the same thing why you should look at what the Gen AI spits out.

正是如此。

Exactly.

它可能极具说服力但完全错误。你必须真正开动脑筋思考。

It could be very convincing, but wrong. And you have to actually, like, put on your thinking cap.

再举个生成式AI的例子——他们测试了几个平台，要求AI为求职者提供起薪谈判建议。假设我是应聘者，提交的简历除姓名外完全一致（姓名可能暗示性别），结果系统给女性名字推荐的起薪谈判值低于男性。

Well, and even looking at the Gen AI example, sorry, one more example is they did some testing of Gen AI platforms when they they were asking the platform to suggest a starting salary negotiation. Say, I'm a job applicant, and here's my resume. Here's the job I'm applying for. The resumes that were submitted were identical except for the name, which could indicate female or male. And the response for the female name was low a lower starting salary negotiation than for the male.

所以即便在

So even in

没错。AI只是从数据中发现了这个代理变量。AI本身不性别歧视，是这个世界存在性别歧视。

Yeah. And the AI found that proxy Yeah. In the data. Because it isn't sexist, the world's sexist. Yeah.

好的，现在正好提出我们的最后一个问题，这与你们讨论的内容相呼应：关于非营利组织帮扶弱势群体的问题，提问者问'您是否认为最需要技术的人群反而最难获得技术？如果是，我们该如何弥合这个鸿沟？'

Right, well this is a good place to have our last really a question here because I think it echoes what you're saying. So someone says, talking about nonprofits and supporting and helping people, do you agree that tech is least accessible to people who need it the most? If yes, how do we bridge that gap?

这确实棘手。以宽带技术为例——最基本的服务接入问题。联邦政府曾提供通讯服务补贴，如今这些补贴被取消，人们不得不在购买食物和支付网费间做艰难抉择。我认为从这种基础层面入手，单是网络可用性就已造成服务获取的不平等。

Yeah, that's a tough one. And I would, one example of that is the broadband technology and that just access to services. There had been subsidies from the federal government for people to have, subsidized service for cell phones or other, connectivity. And those have been cut now, and people are having to make those tough choices between do I buy groceries or do I pay for my Internet bill? But I think just starting at that very fundamental level, it makes a difference in who gets access to services even just based on internet availability.

是的。我想说的是，最需要这些技术工具的人往往最负担不起它们，对吧？好消息是，一旦软件、数据或内容被创造出来，交付它们的边际成本就变得非常非常低。所以重点在于基础设施的建设。

Yeah. I mean, would go so far to say that the people who are most in need of these tech tools are often the least able to afford it. Right? And the good news is the marginal cost to deliver a piece of software or data or content is very, very low once it's been created. And so the focus is on the plumbing.

你有手机吗？能联网吗？家里有宽带吗？但一旦这些条件满足，定价方式很大程度上取决于你的判断。人们可能知道可乐糖浆在撒哈拉以南非洲的价格与北美不同，我们称之为交叉补贴，很多产品都是如此。

Do you have a phone? Does it have a connection? Do you have broadband at your house? But once you get there, in a lot of ways choosing how to price things is really at your discretion. Now people probably know that Coke syrup is priced differently in, you know, Sub Saharan Africa than it is here in, you know, North America, but the same thing is true of a lot we call it cross subsidization.

很可能我们的产品在非洲的售价只有美国的十分之一。在非营利领域，人们对此基本能接受，因为这更多是基于支付能力而非榨取利润。坦白说，如果有人想用我们的技术却完全负担不起，我们很可能会免费提供。作为慈善机构我们可以这样做，但对于有股东或风投压力的企业来说就非常困难。

The odds are quite heavy that if we have a product, we might charge one tenth as much in Africa as we do in The United States. And in the nonprofit sector people actually are kind of okay with that because it's basically based more on ability to pay than on how much we can squeeze you for, right? Know, and frankly if someone wants to use our technology and they can't afford it at all, we're probably going to give it to them for free. But we can do that as a charity. It's very hard to do that if you're a business with shareholders or venture capitalists breathing down your neck.

太棒了。请大家和我一起感谢吉姆和卡米尔带来的精彩节目《科技向善》。这是加州联邦俱乐部国际事务部的项目。非常感谢大家的参与。

Great. Well, listen, please join me in thanking Jim and Camille on a wonderful program, Tech for Good. This is a program of the Commonwealth Club World Affairs of California. Thank you so much for attending.

谢谢，很荣幸。

Thank you. It's a pleasure.

谢谢，谢谢。

Thank you. Thank you.

您正在收听加州联邦俱乐部的节目。您可以在苹果播客、Google Play和Stitcher上收听我们的数千期节目。如果您喜欢我们的内容，请考虑支持我们的工作，帮助我们每年为像您这样的听众带来500场节目。请访问commonwealthclub.org/donate。通过我们的旅行项目，用思维环游世界，探索精彩的国内外目的地。

You've been listening to the Commonwealth Club of California. Hear thousands of our podcasts on Apple Podcasts, Google Play, and Stitcher. If you like what you've heard, please consider supporting our work and help us bring 500 programs a year to listeners like you. Go to commonwealthclub.org slash donate. Think your way around the world with our travel programs to exciting domestic and international destinations.

当您来到湾区时，欢迎亲临我们的现场活动。感谢您的收听与支持。

And when you're in the Bay Area, please join us live at our events. Thank you for listening and for your support.