新闻媒体真的被操控了吗？自从我写下“有效的利他行为其实既无效也不具备真正的利他精神”这一观点以来，已经过去了四年。那个曾经一度衰落的“有效利他主义”运动，在萨姆·班克曼-弗里德因欺诈罪入狱后暂时沉寂下来，如今又卷土重来了。这次，他们让记者们进入新闻编辑部，向我们宣扬人工智能将会如何毁灭人类……媒体的可信度早已岌岌可危了。
Is the news media manipulated? It’s been four years since I wrote that effective altruism is neither effective nor altruistic. The mangy movement, which went into remission when Sam Bankman-Fried went to the big house for fraud, is back. This time it’s placing journalists in newsrooms to tell us how artificial intelligence will kill us all. So much for media’s already fragile credibility.
根据盖洛普（Gallup）的调查，只有三分之一的美国人对媒体抱有“高度信任”或“相当程度的信任”。我认为这种状况其实始于1975年——《周六夜现场》（Saturday Night Live）首次模仿沃尔特·克朗凯特（Walter Cronkite）的新闻播报风格，用幽默的方式削弱了电视新闻的权威性。直到1994年，一位《周末新闻更新》（Weekend Update）的主持人才坦率地承认：“我是诺姆·麦克唐纳（Norm Macdonald）……现在，你们看到的是假新闻。”
According to Gallup, only a third of Americans have a “great deal” or “fair amount” of trust in media. I’m convinced that started in 1975, when “Saturday Night Live” first appropriated the Walter Cronkite-like news desk, and jokes began to water down the perceived credibility of real news from TV news desks. It took until 1994 for an honest assessment from a Weekend Update newscaster: “I’m Norm Macdonald. And now, the fake news.”
五年后，喜剧演员乔恩·斯图尔特（Jon Stewart）在Comedy Central的《每日秀》（The Daily Show）中模仿了正规新闻节目的形式，并使用“头条新闻”（Headlines）风格的字幕来讽刺时事；他的幽默手法很简单：比如在“SNAP福利计划终止的当晚，特朗普发表了一篇充满颓废与享乐主义色彩的演讲”。这种幽默手法其实源自索尔·阿林斯基（Saul Alinsky）的《激进分子的规则》：“嘲笑是人类最有力的武器”——但这种行为根本算不上真正的新闻报道。
Five years later, middling comedian Jon Stewart, on Comedy Central’s “The Daily Show,” copied the legitimate news-set look and used a “Headlines” chyron graphic for his left-leaning news satire. His trick is to go for a quick gag and punch line: “On the very night SNAP benefits ended, Trump threw a Great Gatsby-themed ode to decadence and hedonism.”The laugh line is straight out of Saul Alinsky’s “Rules for Radicals”: “Ridicule is man’s most potent weapon.”But it isn’t news.
乔恩·斯图尔特声称自己不是记者，但他和其他类似的人确实破坏了媒体的可信度。《每日秀》已经获得了26项艾美奖；约翰·奥利弗（John Oliver）在HBO主持的《上周今夜》（Last Week Tonight）同样具有左翼倾向，也经常模仿新闻节目形式，获得了32项艾美奖。这种肤浅、娱乐性的内容总是比有深度的新闻更受观众欢迎。难怪如今近一半的30岁以下美国人都是通过TikTok获取新闻的。
Mr. Stewart says he isn’t a journalist, but he and others like him helped kill media’s credibility. “The Daily Show” has won 26 Emmy awards. John Oliver’s also left-leaning, news-desk-appropriating “Last Week Tonight” on HBO won 32. Quick hits win over substance every time. No wonder almost half of U.S. adults under 30 get their news from TikTok.
我厌倦了看着好莱坞操纵那些赋予公信力的奖项。猜猜今年有多少艾美奖提名或获奖颁给了泰勒·谢里丹，或者颁给了他那些保守倾向的节目《地主》《狮鹫》《麦迪逊》或《权力之城》的演员——甚至多年来颁给了《黄石》？答案是零。“我不想要那些，”厌倦了的谢里丹先生六月对霍华德·斯特恩说。“我不会去领奖。”
I’m tired of watching Hollywood manipulate credibility-bestowing awards. Guess how many Emmy nominations or wins this year went to Taylor Sheridan or actors on his conservative-leaning shows “Landman,” “Lioness,” “The Madison” or “Mayor of Kingstown”—or even “Yellowstone” over the years? That would be none. “I don’t want any of that,” a fed-up Mr. Sheridan told Howard Stern in June. “I won’t go pick it up.”
好了，回到人工智能话题。据报道，Anthropic即将上市。早期投资者包括山姆·班克曼-弗里德的公司FTX（尽管该投资由清算人出售），以及其他有效利他主义爱好者，例如Skype创始人扬·塔林。尽管有效利他主义者发出了“危险，威尔·罗宾逊！”的担忧，但这项投资已取得巨大成功。Anthropic的S-1招股书泄露副本包含80页潜在风险，其中包括其模型对人类构成“灾难性或生存性风险”这一惊人之处。
OK, back to AI. Anthropic is reportedly soon to go public. Early backers included Mr. Bankman-Fried’s company FTX (though the investment was sold by liquidators) and other effective-altruism enthusiasts such as Skype founder Jaan Tallinn. The investment has been a roaring success, despite the “Danger, Will Robinson!” fears of effective altruists. A leaked copy of Anthropic’s S-1 filing contains 80 pages of potential risks, including the doozy that its models pose a “catastrophic or existential risk to humanity.”
首席执行官达里奥·阿莫代九月在联合国表示，“人工智能可能对整个人类构成风险。”尽管阿莫代先生否认自己是有效利他主义者，但Anthropic和OpenAI都有许多有效利他主义者。据报道，Anthropic有5%的股份将进入一个新的“公共利益基金”，用于谁知道什么目的。
CEO Dario Amodei told the United Nations in September that “AI could be a risk to humanity as a whole.”Although Mr. Amodei denies being an effective altruist, there are plenty at both Anthropic and OpenAI. Five percent of Anthropic shares will reportedly go to a new “public benefit fund” for who knows what.
大型有效利他主义基金会Open Philanthropy——该基金会已改名为Coefficient Giving——自2014年以来已拨出70亿美元的资助款。其主要支持者之一是Good Ventures，这是一个由Facebook联合创始人达斯汀·莫斯科维茨和他的妻子卡丽·图纳设立的基金会。Good Ventures网站首页写道：“我们旨在加速进步、减轻痛苦，并减轻对繁荣世界的最大风险。”好吧，非常EA。
The large effective-altruism foundation Open Philanthropy—which has changed its name to Coefficient Giving—has shelled out $7 billion in grants since 2014. One of its large backers is Good Ventures, a foundation set up by Facebook co-founder Dustin Moskovitz and his wife, Cari Tuna. Good Ventures’ website home page reads: “We aim to accelerate progress, reduce suffering, and mitigate the greatest risks to a flourishing world.”OK, very EA.
显然，Coefficient Giving为实现这一目标的策略之一，是资助Tarbell人工智能新闻中心。人工智能新闻报道由人工智能末日论者付费购买。一笔巨额Tarbell资助——该组织计划到2027年底总共支出1000万美元——流向了由人工智能记者凯文·罗斯和凯西·纽顿主持的一档NPR播客。更令人不安的是每年的Tarbell奖学金：“10周的人工智能新闻学基础密集培训”，随后是“9个月的新闻机构实习”。
Apparently, one of Coefficient Giving’s tactics for doing so is funding the Tarbell Center for AI Journalism. The news coverage of AI is paid for by AI doomers. A large Tarbell grant—the organization plans to spend $10 million in total by the end of 2027—went to an NPR podcast by AI journalists Kevin Roose and Casey Newton. More disturbing are the annual Tarbell fellowships: “10 weeks of intensive AI journalism fundamentals training” and then “9 month placements at a newsroom.”
是新闻机构，而非观点媒体。今年的实习单位包括CBS新闻、《洛杉矶时报》、《科学美国人》、CNBC和《南华早报》。
Newsroom, not opinion. This year’s placements include CBS News, the Los Angeles Times, Scientific American, CNBC and the South China Morning Post.
所以，是的，你的人工智能新闻报道可能受到在围绕“人工智能安全”和“生存性风险”议程下训练的记者的影响。这些是Tarbell网站上用于吸引申请的实际话题标签。该中心声称“编辑独立”。当然，我还有一座桥要卖给你。《华盛顿观察家报》对2025年Tarbell学员的研究发现，“人工智能安全和监管是他们报道中出现频率最高的话题之一。”金钱说话。
So yes, your AI news coverage may be tainted by journalists trained in an agenda around “AI safety” and “existential risk.”These are actual topic tags on Tarbell’s website to attract applications. The center claims “editorial independence.”Sure, and I’ve got a bridge to sell you. A Washington Examiner study of the 2025 Tarbell cohort found that “AI safety and regulation were among the topics that appeared most frequently in their coverage.”Money talks.
新闻学校用左翼意识形态灌输未来记者已经够糟糕了——只有约3%的记者是共和党人。但75000美元到135000美元的Tarbell奖学金来操控人工智能新闻，这就太过分了。
Journalism schools that indoctrinate future reporters with leftist ideology are bad enough—only about 3% of journalists are Republicans. But $75,000 to $135,000 Tarbell scholarships to spin AI news is a bridge too far.
我们能相信关于人工智能的新闻报道吗？或者关于许多话题的报道？毫无疑问，左右两派都有大量新闻学奖学金。但投入1000万美元资金来抨击进步？这令人伤心。我一直在寻找类似诺姆·麦克唐纳德的标签，比如“现在，来自有效利他主义包装的新闻。”我会继续寻找。请写信至 kessler@wsj.com。
Can we trust news coverage of AI? Or coverage of many topics? No doubt there are plenty of journalism scholarships on the left and right. But $10 million in funding to bash progress? That hurts. I’ve searched in vain for Norm Macdonald-like labels, “And now, the effective-altruism-spun news.”I’ll keep looking. Write to kessler@wsj.com.