## Background
Published on YouTube by Ai4 on August 27, 2025, this panel discussion explores how artificial intelligence is transforming news discovery, production, and consumption. Moderated by Curtis Sparrer alongside industry leaders Jason Abbruzzese, Justin Hendrix, Kelsey Vlamis, and Nancy Scola, the session examines how newsrooms are adopting generative tools while confronting fundamental ethical, distribution, and business model challenges.
## TL;DR
The panel analyzes the structural disruption AI imposes on digital journalism, particularly the collapse of referral traffic driven by AI search summaries, browser acquisitions, and the rise of synthetic bot spam and 'prompt jacking.' In response, newsrooms are abandoning generic commodity explainers and search-engine aggregation to refocus resources on original accountability reporting, paid subscriber models, and recognizable human journalists. However, the speakers warn that this dynamic risks creating a dangerous two-tiered information ecosystem where reliable reporting is paywalled for subscribers while the broader public is left with an unvetted, AI-manipulated web concentrated in few corporate hands.
## Atmosphere & style
The panel carries a clear-eyed, sobering, and candid tone, frequently punctuated by dry insider humor as veteran journalists confront the existential precarity of their profession. Curtis Sparrer keeps the pacing brisk and engaging, setting up a compelling tension between Justin Hendrix's grim warnings of epistemic collapse and Jason Abbruzzese and Kelsey Vlamis's grounded, pragmatic focus on newsroom survival. Listening feels like sitting inside an unvarnished editorial strategy session where media leaders frankly reckon with the loss of search traffic, platform dependence, and the urgent need to reassert human credibility.
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=4s) Perplexity's bid to acquire Google Chrome underscores an aggressive push by AI search platforms to capture primary web distribution choke points and disintermediate publishers reliant on search referrals.
- [03:04](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=184s) The erosion of digital journalism could revive a pre-twentieth-century landscape of rumors and unvetted pamphleteering unless AI developers are forced to invest in or acquire newsrooms to supply verified ground-truth data.
- [05:04](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=304s) The willingness of readers to pay for niche newsletters and subscriptions enables independent journalists to use generative AI tools as operational force multipliers rather than chasing ad-supported pageviews.
- [07:32](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=452s) Because AI search engines easily absorb generic breaking news aggregation, publishers are compelled to abandon commodity reporting in favor of original investigative enterprise and analytical follow-ups.
- [08:32](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=512s) Practical newsroom AI adoption currently centers on workflow efficiency—most notably automated interview and meeting transcription via tools like Otter.ai—rather than raw story drafting.
- [10:24](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=624s) Features like Google's AI Overviews directly synthesize search results to satisfy user queries on-platform, undermining search engine claims that AI summaries deliver 'high-quality clicks' to publishers.
- [12:20](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=740s) Digital newsrooms are establishing editorial litmus tests to explicitly kill basic informational explainers if a tool like ChatGPT can answer them, redirecting editorial labor to original accountability reporting.
- [15:11](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=911s) Generative Engine Optimization (GEO), or 'prompt jacking,' has emerged as commercial and state actors deliberately format web text and hidden code to manipulate the answers produced by generative answer engines.
- [16:27](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=987s) The proliferation of automated social bot accounts responding to reporting distorts public feedback, creates reporter fatigue, and makes online discourse unreliable for measuring genuine public sentiment.
- [17:30](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=1050s) The information landscape is fragmenting into a two-tiered system where paying subscribers receive verified, human-reported journalism, while the open web remains exposed to low-quality, synthetic manipulation.
- [20:33](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=1233s) Newsrooms risk severe complacency if they assume AI writing quality will remain mediocre, underestimating the capital being invested into rapidly advancing model performance.
- [21:05](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=1265s) Audiences consistently demonstrate higher trust in identifiable human reporters and broadcast personalities than corporate media brands, making reporter-first presentation essential for digital outlets.
- [22:12](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=1332s) Artificial intelligence functions as an engine of power concentration that unites monopolistic tech platforms and state authority, posing profound systemic risks to independent journalism.
## Detailed walkthrough
### [00:04](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=4s) Introductions and Perplexity's Bid for Google Chrome
The panel opens on a lighthearted note with moderator Curtis Sparrer, president of the San Francisco Press Club, promising to keep the discussion lively and quipping that introductions are necessary "in case you need to press charges." Curtis jokes about having crossed over "to the dark side" into PR. He is joined by Kelsey Vlamis, a senior reporter at Business Insider based in Los Angeles who covers business and technology; Jason Abbruzzese, an assistant managing editor for NBC News Digital overseeing technology, science, business, health, and climate; Justin Hendrix, CEO and editor of the nonprofit tech policy outlet Tech Policy Press; and Nancy Scola, a Washington, D.C.-based tech and politics writer and reporter.
To kick off the discussion, Curtis points to breaking news: Perplexity AI making an audacious, multibillion-dollar acquisition offer for Google Chrome well above its typical valuation. Given that journalists increasingly use Perplexity for research and news gathering, Curtis asks what such a move signals for the industry. Justin notes that with a ruling imminent in the Department of Justice's Google search remedies antitrust trial, Perplexity is likely positioning itself early in the event that a forced divestiture of Chrome is ordered, ensuring "somebody needs to have some money on the table when it happens."
Jason, clarifying that his views are strictly his own and do not represent NBC News, frames the bid bluntly: "Look, they're coming for our Google traffic." Search referrals remain the dominant entry point for web audiences across virtually every major media outlet. By capturing Chrome, an AI search entrant like Perplexity could seize that distribution choke point, gain massive user data, and once again disintermediate newsrooms. Kelsey agrees, reminding the audience that while newsrooms are understandably anxious about generative AI upending their business models, journalism was already at the mercy of third-party platforms like Google; generative AI simply raises the stakes.
- **Panel Participants**: Curtis Sparrer (San Francisco Press Club), Kelsey Vlamis (Business Insider), Jason Abbruzzese (NBC News Digital), Justin Hendrix (Tech Policy Press), and Nancy Scola (independent tech/politics reporter).
- **Core Event**: Perplexity AI's aggressive play to purchase Google Chrome amid the ongoing DOJ Google search antitrust trial.
- **The Media Risk**: Disintermediation of publishers who rely heavily on Google Search referrals as their primary audience pipeline.
---
[03:04](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=184s) Bleak Forecasts, the Threat of Dystopian Pamphleteering, and Reasons for Hope
Curtis reads aloud several alarming headlines surfaced by a quick Google search on AI and media: "AI will slash burn the journalism field," "Journalism facing new threats from AI and censorship," and a Pew Research Center study showing Americans overwhelmingly anticipate AI having a negative effect on news. Asking the panel how they process this existential dread, Justin steps in after an awkward silence, joking that journalists are trained to stay quiet until someone breaks and starts talking. On a five-to-ten-year horizon, Justin acknowledges a very grim trajectory where the legacy news industry continues to shrink and atrophy, leaving fewer professional journalists and returning public discourse to a chaotic, pre-twentieth-century landscape dominated by "pamphleteers and rumor and gossip and a stew of nonsense and a kind of general anti-science dystopia."
However, Justin outlines one potential counter-scenario: major AI labs might realize that their models cannot function reliably without ground truth, prompting an "AOL Time Warner" style convergence where AI companies acquire or build newsrooms to bridge the "last mile of information acquisition." Nancy adds another point of tempered optimism: at least the entity attempting to disrupt the browser market is an upstart like Perplexity rather than traditional gatekeepers like Microsoft or the longstanding Google-Facebook digital duopoly, introducing a bit more corporate diversity into the ecosystem.
Jason pushes back against the assumption that top-down corporate newsrooms are the only model that matters. While he hopes major news organizations endure, he sees genuine promise in the unbundling of media through platforms like Substack, where individual journalists and small teams can thrive. As digital consumers become far more accustomed to paying for content online rather than expecting everything to be free, generative AI tools can serve as force multipliers that allow lean operations to produce rigorous, high-value reporting that readers gladly support. Kelsey echoes this sentiment, noting that writing for paying subscribers dramatically raises the editorial standard compared to churning out commodity articles solely designed to harvest ad-driven pageviews.
- **Pew Research Finding**: The American public largely expects artificial intelligence to negatively impact news gathering and journalism.
- **Potential AI Convergence**: Speculation that AI developers may be forced to finance, acquire, or build newsrooms to maintain accurate, real-time data inputs.
- **Audience Monetization**: Consumer willingness to pay for independent newsletters and specialized subscriptions enables individual journalists to operate sustainable businesses using AI tools.
---
[07:32](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=452s) Commodity News vs. Value-Add Journalism and AI Tools in the Newsroom
Turning to the day-to-day mechanics of writing, Kelsey elaborates on the vital distinction between "commodity news" and enterprise journalism. Commodity news refers to the standard, fast-turnaround "click news" that every outlet publishes simultaneously when breaking events occur because anxious audiences are searching for immediate details. In an ecosystem where AI chatbots and search summaries absorb that baseline traffic, newsrooms are forced to rethink their editorial strategy. Rather than racing to push out low-margin aggregation within minutes, reporters can direct their energy toward "second day" follow-ups, deeper investigative angles, and stories that provide unique analysis worth paying for.
NEW MODEL: Differentiated Journalism
[ Breaking News ] ---> [ Fast Timely Update ]
---> [ Skip Generic Explainer ] (Handled by AI)
---> [ Original Reporting & "Day 2" Enterprise ] ---> [ Subscriptions ]
```
When Curtis asks about operational AI tools currently delivering real value in newsrooms, transcription software emerges as an undisputed favorite. Curtis and Nancy both single out Otter.ai as transformative, with Nancy stating that "every young journalist should be introduced to Otter at like day one" to avoid the excruciating, time-wasting chore of manual interview transcription.
Beyond transcription, Nancy highlights the potential of AI to assist with social media copy and headline crafting. In many newsrooms, junior staffers are handed a deeply reported story that took months to investigate and given mere minutes to generate social posts and headlines, often missing critical nuances. Specialized AI tools can draft options that capture the reporting accurately. Nancy notes that while raw ChatGPT struggles to understand what makes a compelling headline, tailored solutions—such as dedicated transcription tools built specifically for local government meetings—are filling key operational niches. The lingering paradox, Nancy notes, is that many of these promising newsroom AI initiatives are funded through grants from Google itself.
- **Commodity vs. Enterprise**: Commodity news satisfies immediate informational queries; enterprise and second-day follow-ups deliver distinct reporting that drives subscriber loyalty.
- **Newsroom Tooling**: Automated transcription (such as Otter.ai) saves massive amounts of reporting time, while specialized headline and government-transcription tools help prevent reporting nuance from getting lost.
- **Platform Dependency**: Even well-intentioned experimental AI tools across newsrooms frequently depend on subsidies and grants from Google News initiatives.
---
[10:24](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=624s) The Crash in Page Views, AI Overviews, and Shifting Editorial Incentives
Curtis addresses the elephant in the room keeping editors awake at night: the precipitous drop in web traffic and page views across digital news publishers. Justin points to a recent analysis published on Tech Policy Press exploring the epistemic consequences of AI search, referencing Google's recent defense of its AI Overviews feature. Google argued that AI summaries do not cannibalize publisher traffic and introduced the concept of "high-quality clicks," suggesting publishers should value better engagement over raw volume. Justin dismisses this spin, arguing that by directly synthesizing answers to informational queries, AI engines eliminate the need for users to click external links across news, civic data, and general web lookups.
Justin warns that this traffic extraction is compounded by the systemic flaws of large language models. Rather than reliable information architecture, platforms are deploying "hallucinatory, sycophantic messes of technology" with minimal safety testing, creating a compounding "miasma of nonsense" at the scale of the open internet. This uncontrolled synthesis stands in direct contrast to the institutional rigor of traditional newsrooms, which have historically exercised extreme caution before placing information into the public square.
Jason connects this distribution crisis directly to editorial decision-making. At NBC News Digital, which operates without a paywall and relies on an advertising-supported business model, traffic incentives historically pushed editors toward high-volume explanatory articles. Today, editorial meetings involve a new litmus test: editors explicitly ask whether ChatGPT or Google Gemini can generate the piece. If the answer is yes, the story is killed. Illustrating this shift, Jason notes that when the Department of Health and Human Services recently announced cuts to mRNA vaccine research funding, NBC reported the breaking news immediately, but chose not to assign an explainer answering "What is mRNA?" While that explainer would have been a mandatory assignment three years ago to capture search volume, it is now abandoned to AI search engines, freeing up reporters to focus strictly on original, scoops-driven accountability reporting.
- **Google's "High-Quality Clicks"**: Tech giants argue AI summaries provide more qualified referrals, but publishers observe an across-the-board reduction in referral visits.
- **Epistemic Degradation**: Justin Hendrix warns that replacing verified sourcing with unverified model outputs creates an unreliable "miasma" across search.
- **The "Can Generative AI Write This?" Test**: NBC News Digital intentionally stops assigning generic informational explainers (e.g., "What is mRNA?"), reallocating editorial labor directly into original, timely reporting.
[15:11](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=911s) Generative Engine Optimization and "Prompt Jacking"
The conversation turns to how the architecture of search and discovery is shifting as journalists and content creators reckon with AI engines. When asked whether writers are actively adjusting their syntax and style to please models like Claude or ChatGPT, Justin Hendrix points out that an entire industry is rapidly coalescing around Generative Engine Optimization (GEO). Hendrix prefers the term "prompt jacking" to describe how actors attempt to game foundational models. Much like traditional search engine optimization (SEO) arose to manipulate Google's algorithms, GEO involves embedding hidden clues, white text, buried code, or specific Markdown structures across websites to skew the answers produced by engines like Perplexity or Google AI Overviews that indiscriminately scrape the public internet.
This tactic is no longer theoretical experimentation. Hendrix emphasizes that nation-states, commercial brands, and politically motivated groups are already pumping out weaponized and optimized text to deliberately manipulate generative outputs and alter what conversational answer engines deliver to the public.
[16:27](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=987s) AI Bots, Misinformation, and the Two-Tiered Information Ecosystem
Addressing the sheer scale of synthetic media and disinformation, panelists discuss how automated interactions are degrading the day-to-day experience of reporting. A panelist recounts tweeting an article about New York City mayoral candidate Zohran Mamdani and instantly receiving a lengthy, synthetic rebuttal from an automated bot account. This relentless flow of machine-generated hostility and feedback wears down reporters, creating a disheartening feeling of throwing serious work into a void only to be shouted down by automated scripts. Kelsey Vlamis observes that while reporters have always relied on online discourse to gauge public sentiment, the proliferation of AI means newsrooms cannot take apparent public reactions at face value without rigorous offline reporting.
```
+------------------------------------------------------------------+
| The Two-Tiered Information Split |
+------------------------------------------------------------------+
| Tier 1: High-Trust Verified Media |
| - Direct subscriber relationships (e.g., NYT, WaPo, NBC News) |
| - Human verification, high editorial friction |
| - Insulated from automated synthetic spam |
+------------------------------------------------------------------+
|
v (Widening Knowledge Gap)
+------------------------------------------------------------------+
| Tier 2: The Open, Unvetted Web |
| - Dependent on search engines, social feeds, AI aggregators |
| - Flooded with GEO prompt-jacking, scrapers, and bot campaigns |
| - Highly vulnerable to automated state & corporate manipulation|
+------------------------------------------------------------------+
```
Taking the role of the panel's optimist, Jason Abbruzzese argues that if the wider web becomes inundated with synthetic garbage, real journalistic credibility will become dramatically more valuable. Outlets like NBC News, *The New York Times*, and *The Washington Post* must clearly articulate their value proposition so audiences understand why they should pay for verified journalism. However, Abbruzzese cautions that this dynamic risks creating a dangerous two-tiered information system where paying subscribers receive verified news, while the broader public relying on open search and social feeds is left vulnerable to mass manipulation.
[19:06](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=1146s) First Amendment Boundaries and State Pressure on Tech and Media
When the panel is asked whether policymakers should regulate AI outputs to curb misinformation, Jason Abbruzzese expresses hesitation, warning that government rules dictating what AI models can or cannot generate quickly run into First Amendment conflicts. He acknowledges that newsrooms are navigating uncharted legal territory.
Justin Hendrix responds with a starkly darker view, arguing that state involvement is already pervasive. He describes the current political climate as one of the most censorious in modern history, marked by heavy-handed government "jawboning," overt political pressure on tech platforms, and transactional, coercive pressure exerted on major media institutions. Rather than intervening to protect truth and journalistic integrity, state power is actively exerting interference across both media and technology companies.
[20:33](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=1233s) Complacency Around AI Quality and the Rise of Human-Centric Branding
The discussion moves to whether journalists are dangerously comforting themselves by assuming that AI-generated writing will remain low quality. Newsrooms risk being blindsided when AI tools inevitably improve, given the massive capital being invested in refining model outputs. Kelsey Vlamis notes that cultivating a strong publication identity offers one defense; readers subscribe to legacy institutions like *The New Yorker* because they want to identify with the publication and what it stands for.
Building on that idea, Jason Abbruzzese highlights the booming popularity of news influencers on TikTok and broadcast figures like Lester Holt. Audiences naturally gravitate toward recognizable human faces, and individual journalists often command higher trust scores than the news organizations employing them. He advocates that digital-first outlets must evolve by putting their reporters directly in front of the audience:
- Audiences seek genuine human connection and clear attribution rather than faceless digital text.
- High-profile broadcast anchors consistently score higher on trust metrics than abstract corporate media brands.
- Digital newsrooms need to push their individual reporters into the spotlight to build personal trust with readers.
[22:12](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=1332s) The Science Fiction Question: Power Concentration and Media Consolidation
Closing with an unscripted question about how severe media consolidation and job losses will become before the industry stabilizes, the panel offers sobering predictions. Panelists agree that the industry has not yet reached bottom, particularly as younger generations have not established habits of paying for journalism or seeking out distinct news brands.
Justin Hendrix delivers a bleak concluding assessment, noting that artificial intelligence is fundamentally a technology of power concentration. The current trajectory merges concentrated corporate control among a handful of tech monopolies with unchecked state power. Between federal contingency planning for rapid domestic deployments and creeping state capitalism in the technology sector, Hendrix warns that the independent press faces unprecedented institutional threats as the broader political and economic landscape rapidly shifts around it.
## Notable quotes
> [00:04](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=4s) in case you need to press charges — Curtis Sparrer
> [00:04](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=4s) somebody needs to have some money on the table when it happens. — Justin Hendrix
> [00:04](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=4s) Look, they're coming for our Google traffic. — Jason Abbruzzese
> [03:04](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=184s) pamphleteers and rumor and gossip and a stew of nonsense and a kind of general anti-science dystopia. — Justin Hendrix
> [03:04](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=184s) last mile of information acquisition. — Justin Hendrix
> [07:32](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=452s) every young journalist should be introduced to Otter at like day one — Nancy Scola
## Mentioned
Curtis Sparrer, Kelsey Vlamis, Jason Abbruzzese, Justin Hendrix, Nancy Scola, Lester Holt, Zohran Mamdani, San Francisco Press Club, Business Insider, NBC News, Tech Policy Press, Perplexity AI, Google, Google Chrome, Department of Justice, Pew Research Center, Microsoft, Substack, Otter.ai, ChatGPT, Google Gemini, Claude, Google AI Overviews, Department of Health and Human Services, The New York Times, The Washington Post, The New Yorker, TikTok
## Video description
Jason Abbruzzese, Assistant Managing Editor, NBC News
Curtis Sparrer, President, San Francisco Press Club
Justin Hendrix, CEO and Editor, Tech Policy Press
Kelsey Vlamis, Senior Reporter, Business Insider
Abstracts
AI is transforming the way news is discovered, produced, and consumed. In this panel, media executives and newsroom leaders discuss how generative tools are streamlining reporting, algorithms are reshaping content distribution, and synthetic media is challenging trust in what we see and hear. We’ll explore how top organizations are using AI to boost efficiency and engagement—while grappling with new ethical, editorial, and regulatory challenges. Whether you run a newsroom or build the tools they use, this session will give you a clear view of where journalism is headed next.
谷歌AI概览分流网络流量，NBC News砍掉解释性报道
生成式搜索引擎正在消灭常规网络引流
专家小组深入分析了人工智能给数字新闻业带来的结构性冲击，特别是AI搜索摘要、浏览器收购，以及合成机器人垃圾内容与“提示词劫持”（prompt jacking）兴起所导致的网络引流暴跌。作为应对，新闻编辑室正在放弃通用且同质化的解释性报道及搜索引擎内容整合，转而将资源重新聚焦于原创监督报道、付费订阅模式以及具有辨识度的真人记者。然而，与会者警告称，这种趋势可能催生危险的两极分化信息生态：可靠的新闻报道被锁定在付费墙后仅供订阅用户阅读，而广大公众面对的则是一个未经核实、受到AI操纵且集中在少数巨头手中的网络环境。
Perplexity 提出收购 Google Chrome 的举动，凸显出 AI 搜索平台争夺网络流量核心分发节点的野心，进一步加剧了依赖搜索引擎推荐流量的新闻机构被边缘化的风险。 (0:04)
新闻编辑室已普遍将 Otter.ai 等自动化转录工具融入日常工作流以提升效率，并主动将采编重心从易被 AI 摘要吸收的同质化快讯，转向更具深度和订阅价值的原创调查报道。 (7:32)
面对 Google AI Overviews 带来的网页访问量下滑，NBC News Digital 等数字媒体设立了新的编辑评估标准，主动砍掉 ChatGPT 即可生成的通识类解释性文章，集中精力投入独家监督报道。 (10:24)
随着生成式引擎优化（GEO，又称“提示词劫持”）的兴起，商业品牌和政治力量正试图通过在网页中植入隐藏文本与特殊排版，刻意操纵生成式搜索答案的输出结果。 (15:11)
Reading summary
Media panelists argue that generative AI answer engines are dismantling the referral-based business model of digital journalism, forcing newsrooms to abandon generic commodity explainers in favor of original, human-centered reporting. However, this disruption threatens to entrench a two-tiered information ecosystem split between paid, verified journalism and an open web vulnerable to manipulation.
Search Disintermediation and Editorial Shifts (10:24)
Search referrals remain the primary traffic driver for digital media, but generative answer engines threaten to seize that distribution choke point. Pointing to Perplexity AI’s acquisition bid for Google Chrome amid antitrust remedies, panelists warn of acute disintermediation. NBC News Digital editor Jason Abbruzzese emphasizes that conversational AI directly extracts publisher audiences. While Google argues that features like AI Overviews produce "high-quality clicks," Tech Policy Press CEO Justin Hendrix counters that direct synthesis eliminates the need for readers to visit news sites altogether, substituting verified reporting with unvetted model outputs.
In response, newsrooms are actively altering editorial priorities. Abbruzzese explains that NBC News Digital now applies an explicit filter: if ChatGPT or Gemini can answer a query, the assignment is dropped. For example, during federal funding cuts to mRNA research, NBC reported the breaking development immediately but abandoned an explainer on "What is mRNA?"—formerly a mandatory assignment to harvest search traffic—allowing reporters to prioritize original enterprise scoops.
Practical Efficiencies and Platform Subsidies in Newsrooms (7:32)
Despite structural threats, newsrooms are realizing tangible operational benefits from targeted AI tools. Automated transcription platforms like Otter.ai have eliminated labor-intensive manual transcription, while specialized tools aid in drafting social copy, headline variations, and processing local government meetings without losing nuanced reporting.
However, reporter Nancy Scola highlights a persistent institutional dependency: many AI experimentation initiatives across newsrooms rely on grants from Google News initiatives. Publishers are thus adopting productivity software backed by the very platforms dismantling their historic distribution mechanisms.
Prompt Jacking and the Two-Tiered Information Split (15:11)
As discovery moves from traditional indexing to conversational answer engines, an optimization sub-industry is rapidly forming. Hendrix warns that commercial brands, political groups, and nation-states are employing Generative Engine Optimization (GEO)—or "prompt jacking"—by embedding hidden code, markdown, and manipulated text across web pages to influence the synthesized answers produced by scrapers and foundational models. Concurrently, automated bot accounts flood social platforms with synthetic pushback, degrading reporters' ability to assess genuine public sentiment.
Abbruzzese cautions that this dynamic risks cementing a two-tiered information ecosystem. High-friction, human-verified reporting will increasingly reside behind paywalls for subscribing audiences, leaving open-web readers reliant on free AI answer engines inundated with automated manipulation and synthetic noise.
Human Attribution and Expanding Power Concentration (20:33)
Panelists warn journalists against complacency regarding the current limitations of AI text quality, noting that sustained capital investment will inevitably improve model fluency. Business Insider's Kelsey Vlamis and Abbruzzese suggest that outlets fortify their value proposition by elevating individual reporters. Because audiences frequently trust recognizable human figures over abstract corporate news brands, digital publishers must adopt influencer-style transparency and direct attribution to build audience loyalty.
Nevertheless, the long-term outlook remains fraught. Hendrix concludes that artificial intelligence fundamentally operates as a tool of power concentration, consolidating influence among a handful of tech monopolies while intersecting with increasing state pressure on media institutions, leaving the future of an independent press precarious.
While independent subscription models and direct audience engagement offer survival mechanisms for individual journalists, the panel leaves unresolved how the broader public will access credible information as generative answer engines degrade the open web.
## Background
该内容由 YouTube 频道 Ai4 于 2025 年 8 月 27 日发布，视频标题为《AI 如何改变新闻业》（How AI Is Changing the News）。根据页面描述，本场研讨会汇集了 NBC News、旧金山新闻俱乐部、Tech Policy Press 和 Business Insider 的媒体高管与资深记者，共同探讨生成式 AI 工具如何简化新闻报道流程、算法对内容分发格局的重塑，以及合成媒体对信息信任带来的挑战与伦理监管困境。
## TL;DR
与会嘉宾深入探讨了生成式 AI 与 AI 概览（AI Overviews）对传统搜索引擎导流模式的冲击，指出新闻媒体正面临严重的“去中间化”危机与流量下滑风险。面对 AI 垃圾信息的泛滥与生成式优化（GEO）等新型内容操纵手段，新闻机构正重塑采编激励机制，逐步放弃可被 AI 轻易替代的“商品化新闻”，将核心资源全面倾斜于原创深度调查与高时效报道。与此同时，Substack 等个人新闻业的兴起与读者付费订阅习惯的转型，正在推动媒体行业形成高壁垒权威品牌与免费虚假信息交织的“双层化”信息体系。
## Atmosphere & style
整场研讨会气氛专业严肃、理性而充满警醒，嘉宾们既有对行业生存危机的深切担忧，也不乏对新闻本质回归的敏锐洞察。旧金山新闻俱乐部主席的幽默开场为讨论奠定了坦诚交流的基调，随后嘉宾们结合一线采编与策略实操，展开了环环相扣的深度碰撞。对话节奏紧凑、针砭时弊，展现出新闻人在技术巨头裹挟与信息迷雾冲击下，对坚守新闻专业主义与重塑行业伦理的深刻思考。
## Key takeaways
- [01:08](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=68s) Perplexity 竞购 Chrome 凸显出生成式 AI 初创公司争夺网络浏览器关键入口的野心，旨在重塑搜索与信息分发格局。
- [01:08](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=68s) AI 生成式搜索介入使得新闻媒体再次面临“去中间化”危机，传统新闻机构依赖搜索引擎导流的流量分配机制被严重切断。
- [02:43](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=163s) 传统新闻业面临生存威胁，未来 5 到 10 年专业记者数量可能锐减，甚至导致社会倒退回充满虚假信息的“小册子作者时代”。
- [02:43](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=163s) 科技公司为保证 AI 内容准确性，未来可能直接收购或自建新闻团队，同时新兴 AI 公司的崛起也打破了传统科技巨头的绝对垄断。
- [05:43](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=343s) Substack 等个人新闻业与微型团队借力 AI 展现出强劲活力，行业商业模式正从依赖页面浏览量转向吸引读者付费订阅。
- [07:34](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=454s) 新闻行业加速分化为同质化的“商品化新闻”与需要深度采编的“独家报道”，媒体逐步放弃对低质流量文的投入。
- [08:22](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=502s) Otter.ai 等 AI 工具已广泛应用于采编转录与摘要提炼，但通用大模型在把握报道微妙语境及撰写优质标题上仍有局限。
- [10:25](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=625s) Google 的“AI 概览”直接在搜索顶部汇总信息，导致新闻网站搜索引擎流量出现断崖式下滑，并引发网络信息垃圾化的风险。
- [13:18](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=798s) 流量下滑倒逼新闻编辑部改变激励机制，停止分配 AI 可轻松解答的解释性任务，将核心资源集中于原创突发报道与独家调查。
- [15:11](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=911s) 围绕生成式 AI 引擎的“生成式优化”（GEO）行业逐步形成，国家级行为体与商业机构正利用特制格式数据试图操纵 AI 输出。
- [16:27](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=987s) AI 大幅降低了高逼真度虚假内容的制作门槛，自动化 Bot 针对记者的攻击推高了验证成本，促使权威媒体靠建立高壁垒品牌维系信任。
- [16:27](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=987s) AI 时代的社会信息生态正在形成“双层化”割裂：愿意付费的群体获得可信信息，依赖免费平台的群体沦为虚假信息的受害者。
- [19:06](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=1146s) 在讨论立法者对 AI 生成内容进行政策干预时，面临着第一修正案（言论自由）等法律界限以及政府政治干预的复杂考量。
## Detailed walkthrough
### [00:04](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=4s) 研讨会嘉宾自我介绍与行业背景
会议主持人和与会嘉宾依次进行了自我介绍。旧金山新闻俱乐部（San Francisco Press Club）主席柯蒂斯·斯帕勒（Curtis Sparrer）开场打趣称，自己虽然投身于公关行业这个“黑暗阵营”，但仍致力于推动媒体交流。随后，Business Insider 驻洛杉矶的资深科技与商业记者凯尔西·弗拉米斯（Kelsey Vlamis）、NBC News Digital 助理总编辑杰森·阿布鲁泽塞（Jason Abbruzzese，负责科技、科学、商业与气候报道）以及 Tech Policy Press 首席执行官兼总编辑贾斯汀·亨德里克斯（Justin Hendrix）分别介绍了各自的职责领域。
### [01:08](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=68s) Perplexity 竞购 Chrome：搜索引擎与媒体流量的再中间化
研讨会首先聚焦于近期发生的重磅事件——人工智能搜索初创公司 Perplexity 对 Google Chrome 浏览器发起的收购邀约。主持人与嘉宾指出，这发生在针对谷歌搜索垄断的司法裁决即将落地的敏感时期，可能存在抢先布局剥离资产的考量。从媒体运营角度来看，Google 搜索长期以来是各大新闻机构最核心的流量来源；Perplexity 意图掌控浏览器这一关键入口，旨在借此捕获海量用户与数据，从而在信息分发格局中占据主导地位。
阿布鲁泽塞指出，这种趋势本质上是新闻媒体被“再次去中间化”（Disintermediated）的过程。新闻的发现与获取管道再次被第三方平台截留，传统媒体不得不面对流量分配权进一步被稀释的现实。
- **谷歌搜索垄断裁决**：行业正密切关注即将出炉的搜索补救措施裁决，Perplexity 的高价竞购被视为预判谷歌可能被迫剥离 Chrome 的前瞻动作。
- **流量去中间化风险**：新闻机构长期依赖搜索引擎导流，生成式 AI 工具的介入切断了用户直接点击进入新闻原址的路径。
### [02:43](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=163s) AI 对新闻业的生存威胁与历史周期演变
针对媒体行业面对 AI 冲击的焦虑，亨德里克斯对新闻业的未来前景提出了深切担忧。他预测在未来 5 到 10 年内，传统新闻机构可能持续萎缩，专业记者数量锐减，社会甚至可能倒退回类似 20 世纪之前的“小册子作者（Pamphleteers）时代”——充斥着谣言、八卦、反科学言论与大量虚假废话。
不过，亨德里克斯也提出了另一种可能的演变路径：如果科技公司为了确保其 AI 系统输出内容的准确性与实时性，可能会出现类似当年 AOL 与时代华纳合并的融合时刻，即 AI 巨头选择直接收购或组建自己的新闻团队，以解决信息采集的“最后一公里”问题。此外，相比过去 Google 与 Facebook 组成的“数字双头垄断（Digital Duopoly）”，现阶段 Perplexity 等新型科技公司的崛起虽然增加了不确定性，但也打破了绝对垄断，为市场带来了一丝多样性。
### [05:43](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=343s) 个人新闻业崛起与订阅制付费模式转型
新闻消费格局的演变正在重塑媒体的商业模式。传统上由大型集团自上而下统揽的中心化报道体系，正在被以 Substack 为代表的个人新闻业（Individual Journalism）与微型新闻团队所补充。得益于生成式 AI 工具的赋能，极少数记者甚至单个独立创作者也能高效生产出极高品质的深度新闻报道。
与此同时，互联网用户的付费意愿发生了根本性转变。从最初的“免费互联网时代”到后来的信用卡在线支付习惯普及，如今读者越来越愿意为优质的独家内容付费。弗拉米斯强调，这种从“博取页面浏览量（Page Views）”向“吸引订阅付费”的转型，极大地抬高了新闻采编的门槛，激励记者摆脱低质流量文，专注于创作具有长期价值与读者忠诚度的深度报道。
### [07:34](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=454s) “商品化新闻”与“深度独家内容”的分流
研讨会深度探讨了两种截然不同的大众新闻类型：
1. **商品化新闻（Commodity News / Click News）**：即各大媒体争相报道的热点追逐类文章。其核心在于快速上线以捕获即时搜索流量，但同质化严重。
2. **深度独家内容（Second-day / Deep Stories）**：关注事件背后的推演、后续影响与实质洞察，需要花费更多时间采编，是吸引用户付费的关键。
在 AI 流量导流能力下滑的背景下，媒体机构正在逐步放弃对商品化新闻的盲目投入，转而将核心资源集中于生产 AI 无法轻易替代的深挖型报道。
### [08:22](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=502s) AI 工具在新闻采编中的具体应用与局限
谈及新闻工作流中的工具应用，嘉宾们分享了生成式 AI 在生产端带来的效率提升与潜在风险。
```
[采访录音 / 原始素材]
│
▼
[ Otter.ai 自动转录 ] ──► 节省人工逐字听写时间
│
▼
[ 专用生成式 AI 工具 ] ──► 提炼政府会议摘要 / 辅助生成社交媒体标题草稿
│
▼
[ 记者人工审核与润色 ] ──► 输出最终高品质报道
```
尽管转录工具（如 Otter.ai）和自动化摘要大幅解放了记者的时间，但通用大语言模型（如 ChatGPT）在撰写具有新闻价值的爆款标题或把握报道微妙语境方面仍存在局限。此外，许多新闻创新项目（如会议记录分析工具 Minutes）在资金与技术上高度依赖谷歌等巨头的补贴，使得新闻业对科技巨头的依附关系难以彻底切断。
### [10:25](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=625s) AI 概览（AI Overviews）对搜索流量的毁灭性打击
与会嘉宾直面了当前令所有线上新闻网站失眠的危机——搜索引擎流量的断崖式下滑。亨德里克斯引用了最近针对 Google “AI 概览（AI Overviews）”的研究，揭示了生成式搜索对互联网信息生态的重塑。
谷歌通过 AI 概览直接在搜索结果顶部对信息进行综合汇总，使得绝大多数用户不再需要点击进入原始新闻链接。即便谷歌提出了所谓“高品质点击（High-quality Clicks）”的概念来进行辩解，实际情况却是生成式 AI 正在切断媒体的流量根基。更严重的是，这些 AI 引擎本身存在幻觉（Hallucinations）、谄媚性（Sycophancy）等固有缺陷，未经充分测试便大规模投放给公众，正在将互联网变为由“AI 垃圾信息循环互引”构成的虚假迷雾（Miasma of Nonsense）。
### [13:18](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=798s) 新闻采编激励机制的根本性重塑
面对流量下滑与 AI 的侵蚀，依靠广告变行、无付费墙（Paywall）的传统网媒被迫重新审视新闻选题策略。阿布鲁泽塞以 NBC News 的实际操作为例，说明了 AI 如何重塑新闻编辑部的工作流程与选题逻辑：
- **停止生产 AI 可替代内容**：对于诸如“什么是 mRNA（What is mRNA?）”这类解释性知识点，过去在重大新闻突发时会专门派发任务以捕获搜索流量，但如今大模型能完美回答，编辑部已全面停止分配此类撰稿任务。
- **资源向独家与实效倾斜**：编辑部不再将精力消耗在简单的信息汇编（Information Aggregation）上，而是将节省下来的时间与人力资源全部投入到“第一时间突发报道”与“原创深度调查”中。
这种激励机制的被迫重组，正促使新闻业重返 journalism 的本质——专注于生产高时效、高质素及具有社会影响力的原创报道。
<ElicitationsGroup message="想要深入探索 AI 与新闻业变革的相关话题？">
<Elicitation label="分析生成式 AI 对传统新闻媒体商业模式的深远影响" query="详细分析生成式 AI 和搜索引擎 AI 概览（AI Overviews）对传统新闻媒体商业模式与流量变现的深远影响。"/>
<Elicitation label="对比解析独立新闻（如 Substack）与大型新闻机构的未来发展" query="对比解析在 AI 时代下，Substack 等个人/微型新闻团队与传统大型新闻机构各自的竞争优势与生存策略。"/>
</ElicitationsGroup>
### [15:11](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=911s) 针对 AI 引擎的“生成式优化”与 SEO 新变革
随着生成式 AI 的普及，媒体与内容创作者正在面临类似早期搜索引擎优化（SEO）的全新转变。针对“是否有人为了迎合 Claude 或 ChatGPT 等 AI 引擎而改变写作习惯”的提问，专家们指出，一个全新的行业正围绕生成式引擎优化（Generative Engine Optimization，简称为 GEO，部分与会者戏称为“Prompt Jacking”/提示词劫持）逐步形成。
这种机制的原理在于，通过在网站上散播海量特定格式的信息——包括人类不可见的白色文本、特制代码或 Markdown 格式——来迎合 Perplexity、Google 等无差别抓取全网数据的 AI 工具，从而人为干预或重塑 AI 输出的内容。
- **多方博弈**：目前，已有国家级行为体（Nation states）、商业品牌以及带有政治意图的组织利用这一手段，试图操纵 AI 模型生成的内容方向。
---
### [16:27](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=987s) 虚假信息危机与新闻信任度的“双层化”割裂
AI 技术大幅降低了高逼真度虚假内容的制作门槛，为国家级力量和利益集团提供了强有力的操纵工具。在日常工作中，新闻记者和编辑们已经感受到了这一趋势带来的负面效应，例如在社交平台上发布新闻后，会遭遇 AI 自动回复机器人（Bot）针对性的长篇大论攻击。这种源源不断的自动化质疑不仅对记者造成心理消耗，也极大推高了验证网民真实反应的成本。
面对这一困境，传统媒体提出了应对策略与隐忧：
1. **建立高壁垒品牌**：在网络充斥 AI 垃圾信息的环境下，权威新闻机构（如 NBC News、《纽约时报》、《华盛顿邮报》）需要重新确立核心价值主张，让读者明确付费订阅的理由。
2. **“双层化”信息体系（Two-tiered information system）**：愿意为优质深度报道付费的群体能够获得真实可信的信息；而依赖免费平台与社交网络的人群，则极易沦为 AI 虚假信息和内容操纵的受害者。
---
### [19:06](https://www.youtube.com/watch?v=6uIFAMdRy-s&t=1146s) 政策监管困境与政府对媒体及科技巨头的干预
针对立法者是否应出台政策干预 AI 生成内容的讨论，嘉宾们表达了对第一修正案（言论自由）等法律界限的顾虑，同时指出了当前美国政治与政策环境对新闻自由的深刻影响。
批评者指出，当前美国联邦政府正表现出前所未有的审查与干预倾向。政府不仅直接打
## Mentioned
Jason Abbruzzese, Curtis Sparrer, Justin Hendrix, Kelsey Vlamis, NBC News, San Francisco Press Club, Tech Policy Press, Business Insider, Perplexity, Google Chrome, Google, Substack, Otter.ai, Minutes, ChatGPT, Claude
## Video description
Jason Abbruzzese, Assistant Managing Editor, NBC News
Curtis Sparrer, President, San Francisco Press Club
Justin Hendrix, CEO and Editor, Tech Policy Press
Kelsey Vlamis, Senior Reporter, Business Insider
Abstracts
AI is transforming the way news is discovered, produced, and consumed. In this panel, media executives and newsroom leaders discuss how generative tools are streamlining reporting, algorithms are reshaping content distribution, and synthetic media is challenging trust in what we see and hear. We’ll explore how top organizations are using AI to boost efficiency and engagement—while grappling with new ethical, editorial, and regulatory challenges. Whether you run a newsroom or build the tools they use, this session will give you a clear view of where journalism is headed next.
[Source](https://www.youtube.com/watch?v=6uIFAMdRy-s)
Source quotes
[0:04] Curtis Sparrer: in case you need to press charges
[0:04] Justin Hendrix: somebody needs to have some money on the table when it happens.
[0:04] Jason Abbruzzese: Look, they're coming for our Google traffic.
[3:04] Justin Hendrix: pamphleteers and rumor and gossip and a stew of nonsense and a kind of general anti-science dystopia.
[3:04] Justin Hendrix: last mile of information acquisition.
[7:32] Nancy Scola: every young journalist should be introduced to Otter at like day one
[10:24] Justin Hendrix: high-quality clicks
[10:24] Justin Hendrix: hallucinatory, sycophantic messes of technology
[10:24] Justin Hendrix: miasma of nonsense
[10:24] Jason Abbruzzese: What is mRNA?
[15:11] Justin Hendrix: prompt jacking
Original transcript
[0:04] Thanks for joining us. I promise we'll
[0:06] try to be lively and fun. If not, I'll
[0:07] interrupt. But we want to start off with
[0:09] introductions just so you have a sense
[0:11] of who we are and in case you need to
[0:13] press charges. So, my name is Curtis uh
[0:16] Sparer. Actually, I am the president of
[0:18] the San Francisco Press Club. I did
[0:20] switch over to the dark side and went to
[0:22] PR bills. Uh so, that's the short short
[0:26] version about me. What about you?
[0:29] I'm Kelsey Blamis. I'm a senior reporter
[0:31] at Business Insider where I cover
[0:33] business and tech news. Um, I'm based in
[0:35] LA and I've been at Business Insider for
[0:37] almost 5 years. Uh,
[0:39] >> I'm Jason Bruises. I'm an assistant
[0:41] managing editor uh for NBC News Digital.
[0:43] I oversee our technology, our science,
[0:45] our business, health, science, and
[0:47] climate coverage.
[0:48] >> I'm Justin Hendricks. I'm a CEO and
[0:50] editor of a nonprofit uh entity called
[0:52] Tech Policy Press, which sort of does
[0:54] what it says on the package. We cover
[0:56] tech policy. Uh Nancy Skull. I'm a tech
[0:58] and politics uh writer reporter based in
[1:00] DC.
[1:00] >> And Nancy, I mean like everything you
[1:02] said probably took all our lines away.
[1:04] We have nothing, no content. So thank
[1:06] you. But it was a really good session.
[1:08] But I actually wanted to start off with
[1:10] a little bit of a breaker. Uh some of
[1:12] you may have seen that Perplexity made a
[1:15] big offer for Google Chrome, way over
[1:18] what they're valued. And I kind of think
[1:21] this is kind of part and parcel of what
[1:23] we're discussing because I don't I know
[1:24] a lot of us use perplexity for, you
[1:27] know, news monitoring and gathering. So
[1:29] anyone, what do you think this means?
[1:32] >> Well, I mean, we're expecting a ruling
[1:34] any day now in the Google uh search
[1:36] remedies trial. Um, and so maybe they're
[1:40] trying to kind of preemptively hope that
[1:42] uh devestature is required and uh
[1:45] somebody needs to have some money on the
[1:47] table when it happens.
[1:50] That's it.
[1:51] >> Well, I I have to preface this by saying
[1:53] these are my opinions, not the opinions
[1:54] of NBC News. Uh, look, they're coming
[1:57] they're coming for our Google traffic,
[1:58] right?
[1:59] >> Like most of the media that uh, you
[2:01] know, I think relies on Google traffic.
[2:04] Google search traffic is still, you
[2:06] know, a top refer for just about every
[2:08] uh, media organization out there. And
[2:10] Perplexity sees that and says, "Look,
[2:13] this is a great entry point for people
[2:15] seeking information. If we can uh
[2:17] harness that uh and take advantage of
[2:19] it, we're going to have a ton of users,
[2:22] you know, and a ton of data and we're
[2:23] going to be uh a bigger and better
[2:25] player and more power to them. Um but it
[2:27] does mean that, you know, uh journalists
[2:30] and in news media are getting, you know,
[2:32] disintermediated yet again. Somebody is
[2:34] coming in between, you know, where
[2:36] people end up finding news and saying
[2:38] like, uh I think that you guys have had
[2:41] that long enough.
[2:43] >> Yeah. Speaking Oh, you were going to say
[2:44] something? Yeah, I would add that it's
[2:45] also a reminder that while there is
[2:47] understandable concern about how AI is
[2:49] going to disrupt our traffic, we are
[2:51] already being impacted by third parties
[2:53] like Google, like it's it's not
[2:55] necessarily new that this third party is
[2:57] now coming in and it's disrupting our
[2:58] business model. But I mean it is a
[3:00] reminder that I the stakes are kind of
[3:02] rising. Yes, the stakes are rising. In
[3:05] fact, I did a little research which is
[3:07] called just looking at my cell phone
[3:09] kind of during a you know important part
[3:11] Jason was making and I looked and I
[3:13] Googled AI and journalism and a couple
[3:16] headlines uh AI will slash burn the
[3:19] journalism field journalism facing new
[3:21] threats from AI and censorship. American
[3:23] largely foresee AI is having negative
[3:25] effects on news and journalists from the
[3:27] Pew Research Center. And you know, gosh,
[3:31] we haven't thrown oursel off a cliff
[3:32] yet, but how does this kind of impact
[3:34] you and what does it make you think?
[3:36] >> Nancy, come on. You just gave a whole
[3:38] talk on this.
[3:38] >> Um, I feel like I'm jumping in to fill
[3:41] the silence, which is always a terrible
[3:42] thing to do when someone's asking you
[3:44] questions. That's what they teach
[3:46] journalists to do. Keep quiet and
[3:48] someone will begin to talk. Um, what
[3:51] does it make me feel? What does it make
[3:52] me think? Um I suspect that uh to my
[3:57] question to Nancy at the end of her talk
[3:59] um 5 to 10 years from now uh the
[4:02] existing industry has continued to
[4:04] shrink uh atrophy. Um there are fewer
[4:07] professional journalists in the United
[4:09] States and elsewhere in the world. Um we
[4:12] uh are effectively uh in a kind of you
[4:16] know back to a kind of pre 20th century
[4:19] uh age of of pamphleteers and uh rumor
[4:22] and gossip and a stew of nonsense and uh
[4:25] you know kind of uh general anti-science
[4:28] uh dystopia. Um, I'll make one caveat
[4:32] and that is uh the whole thing might get
[4:35] turned around if in fact tech firms
[4:38] decide that in order to make their AI
[4:40] systems work, they need to hire uh some
[4:44] humans to to help make certain that the
[4:47] information they have is current and
[4:49] accurate. Um, so I could see a kind of
[4:52] convergence moment maybe something
[4:53] similar to the ideas that were around in
[4:56] the as you know around AOL Time Warner
[4:59] merger time where uh the AI firms either
[5:02] acquire or build their own newsrooms in
[5:05] order to satisfy the last mile of
[5:07] information acquisition they need to do.
[5:10] I guess the one thing I might add is
[5:11] that maybe it's one note of optimism
[5:13] might be that it's perplexity talking
[5:15] about which is news you just broke to me
[5:17] but buying uh Google Chrome and not
[5:19] Microsoft or the few companies that in
[5:21] the past had been the really the sort of
[5:23] gatekeepers of the internet at least
[5:25] there's a little bit of flourishing of
[5:26] uh of tech companies so that we're not
[5:28] so dependent it was all you know the
[5:30] digital duopoly was always Google and
[5:32] Facebook and we uh were just sort of
[5:34] subject to their whims at least there's
[5:35] a a handful more companies that that now
[5:38] were subject uh to their whims um But a
[5:40] little bit more diversity in the field
[5:41] might be a good thing.
[5:43] >> Yeah. And I I I also think it very much
[5:44] begs a question of like what the future
[5:46] of of journalism will look like, which
[5:48] is the title of this panel. But, you
[5:50] know, I think we did get used to a world
[5:52] in which it was very corporate. Um, it
[5:54] was very top down. Um, you know, either
[5:56] it was your city newspaper or it was a
[5:58] national news. Um, you were always
[6:00] working for a company. I do think the
[6:02] rise of Substack and the rise of kind of
[6:04] individual journalism uh is very notable
[6:06] and I think NY's parts of NY's uh
[6:08] presentation actually make me very
[6:10] optimistic that you know a small number
[6:12] of journalists will be able to generate
[6:14] like really good highquality news and I
[6:18] think that that dovetales with the
[6:19] reality that people are getting more and
[6:21] more comfortable paying for things
[6:22] online. There was a long time where
[6:23] everything on the internet was free. It
[6:25] was just you couldn't even pay for stuff
[6:26] if you wanted to. That hadn't been built
[6:28] yet. Then we had the era of putting your
[6:30] credit card in this And the internet is
[6:31] insane. People got comfortable with
[6:33] that. You know, now you see people, how
[6:35] many substacks, you know, do does the
[6:37] average person subscribe to? I don't
[6:39] think that many yet. But I also think
[6:40] that, you know, you're seeing at a time
[6:42] where AI tools are also providing a way
[6:44] for individuals and small groups to
[6:46] provide a lot of value to people who are
[6:48] getting increasingly comfortable paying
[6:50] for things on the internet. And that
[6:51] gives me some hope that the future of of
[6:53] news on some level uh you know I speak
[6:56] for working for a large national news
[6:58] organization that I hope doesn't go away
[7:00] anytime soon. But the optimism in me is
[7:02] very much keyed into what uh small
[7:04] groups and individuals are able to
[7:06] accomplish.
[7:07] >> And I would add to that that if people
[7:10] are willing to pay for news and we know
[7:11] that they are like Jason said the rise
[7:12] of Substack the kind of work you're
[7:15] producing when people are paying for it
[7:17] really raises the bar as opposed to when
[7:18] you're just trying to get page views.
[7:20] Um, so like as a reporter, I'm actually
[7:23] really excited for this shift. Like I
[7:24] don't want to write stories just to get
[7:26] paid views. I want to write stories that
[7:27] people want to pay for, they want to be
[7:28] loyal to. So in that respect, the
[7:30] optimist in me is a little bit excited.
[7:32] >> Let's just talk quickly about the
[7:34] difference between those two types of
[7:35] stories. We we did at lunch at the very
[7:37] well- attended prep session that we had
[7:39] for this panel. Uh
[7:43] what's the difference?
[7:44] >> So I a term that gets used sometimes is
[7:47] commodity news. And I think of that as
[7:49] click news almost. It's like everyone's
[7:52] writing it, but you kind of need to
[7:53] write it because everyone is reading it.
[7:54] Like readers, everyone is anxious or
[7:57] they're trying to find information about
[7:58] this specific topic right now. So, we
[7:59] need to put up a story really quickly
[8:01] about this. We want to make sure we
[8:02] harness that traffic. But in an age of
[8:05] AI where maybe that traffic isn't as
[8:07] relevant, we can focus on, okay, well,
[8:09] how do we push the story forward? What's
[8:10] the second day story? What's the deeper
[8:12] story here? Like, how do we produce
[8:13] higher quality content that maybe we
[8:15] don't put out today? We maybe don't get
[8:16] those clicks today. But if we put out a
[8:18] really great story tomorrow, someone
[8:19] will pay for us.
[8:20] >> Okay. And Nancy, you went through a lot
[8:22] of the different tools that you were
[8:25] using or seeing that were beneficial.
[8:27] I'm wondering for the rest of us who are
[8:29] watching along thinking, "Damn it, I
[8:30] wish I'd said that or saved that for
[8:31] this session." Any other tools or
[8:34] operational impacts you're seeing from
[8:37] AI being used in the newsroom? I'm a
[8:39] huge fan of taking notes by Otter, for
[8:41] example, and crashing that into a story.
[8:43] What are you guys using? would I I think
[8:47] every young journalist should be
[8:48] introduced to Otter at like day one of
[8:50] >> this session is not sponsored by Otter
[8:52] by Otter. It's the best.
[8:54] >> Um it just saves you I mean I am old
[8:56] enough to have remembered hand to
[8:57] transcribing interviews um and you know
[9:00] scrolling through the transcript to
[9:01] figure out the what nugget was going to
[9:03] be important. Um it was horrible, it was
[9:04] painful, it was a wast giant waste of
[9:06] time. So that's um that's tremendous. So
[9:08] the question is other tools that
[9:09] >> other tools yes
[9:11] >> um I mean I think the the idea of like
[9:14] the the social media aspect of this I
[9:16] think very often I don't know your
[9:17] experience in news rooms very often it's
[9:19] a you know very new um journalists in
[9:22] the newsroom who's charged with writing
[9:24] social media writing headlines for
[9:25] stories that we might put months into
[9:27] and they spend a couple minutes um
[9:29] coming up with a headline and social
[9:30] media around it that often doesn't
[9:32] reflect the story the sort of nuances of
[9:33] the story and that's not their fault
[9:34] right they're given a few minutes to
[9:35] kind of come up with that um so I think
[9:37] using some of those tools that I talked
[9:39] about um to um come up with uh better
[9:42] drafts of um that content I think uh
[9:46] could be really useful and I think
[9:47] specialized tools I think that uh you
[9:49] know using chat GPT for this stuff
[9:51] chatpt PT doesn't know what makes a
[9:52] great headline um but people are working
[9:54] on building I mentioned that minutes
[9:55] tool which is transcription but
[9:57] specifically um catered around
[9:59] government meetings we could have some
[10:01] of these AI generative AI tools that are
[10:02] really you know what is what makes a
[10:04] great headline give me some rough
[10:05] content to work with so I'm hopeful that
[10:07] people are going to keep innovating in
[10:08] that space. One challenge is a lot of
[10:09] the projects I talked about that I think
[10:11] are neat are sponsored by Google, right?
[10:13] Still, we get back to the fact that
[10:14] we're still dependent on these
[10:15] companies. They're just kind of giving
[10:16] us a little bit of subsidy in the, you
[10:18] know, as part of their news initiatives.
[10:19] >> Oh, you we'll take it. We'll take it.
[10:21] >> We'll take it. But like, you know, I
[10:22] mean, do you feel good about like
[10:24] >> one thing I don't feel good about, just
[10:26] to kind of segue, is we're seeing a lot
[10:29] of online news sites lose page views.
[10:33] We're we're seeing a crash. truly that
[10:36] keeps me up at night and I'm wondering
[10:38] what you all think is causing that and
[10:41] what we can do besides just write better
[10:44] to you know solve that.
[10:47] We have a piece on our site, I think
[10:49] it's still on featured on the kind of
[10:51] third down that talks a lot about uh
[10:53] Google AI overviews. Um it's by a couple
[10:56] of researchers who are kind of asking
[10:58] this question about um the kind of
[11:00] epistemic effects of AI search.
[11:03] Uh, one of the things that it does is it
[11:05] it it um addresses what I think you
[11:07] referred to, which is Google's response,
[11:10] uh, just last week, I believe, um, this
[11:12] blog post that they had about, uh, the
[11:14] concerns that have been raised about
[11:16] whether AI overviews is essentially, um,
[11:18] stealing traffic from from sites. And of
[11:21] course, Google has reassured us, as you
[11:23] mentioned, um, that, you know, that's
[11:25] not happening quite the way that has
[11:27] been portrayed. And they even introduced
[11:29] this new idea of I think highquality
[11:31] clicks. Uh so there's a difference
[11:33] between clicks and high quality clicks.
[11:35] Um and of course we should all be going
[11:37] for uh the latter, right? Um so uh you
[11:41] know I think it seems fairly clear what
[11:44] is happening. um that it's not just
[11:47] Google but a variety of uh AI services
[11:52] uh essentially are synthesizing
[11:54] information in a way that makes it
[11:56] unnecessary for individuals to click on
[11:58] through um and that is you know
[12:01] obviating the need I think that's not
[12:02] just happening with news it's also
[12:04] happening with uh government information
[12:06] it's happening with uh you know any form
[12:08] of of uh query that I might do around
[12:11] information on the internet um the
[12:13] problem becomes you know uh that all of
[12:16] these services are flawed, deeply
[12:20] flawed, hallucinatory, sycopantic,
[12:24] uh you know, uh messes of technology
[12:28] that have been foisted on the public,
[12:30] often with extraordinarily little
[12:32] concern for testing or potential impact
[12:35] on uh any of the broader concerns we
[12:38] might have as a society. Um, so we're in
[12:40] this situation now where essentially now
[12:43] you've got u various services building
[12:46] on top of the muck generated by these AI
[12:50] uh engines. And so we're we're
[12:52] eventually think about this happening at
[12:54] the scale of the internet. Um we're
[12:56] building a kind of uh miasma of nonsense
[13:00] um which opposite.
[13:02] >> What's that?
[13:02] >> I use a miasma.
[13:04] >> Oh, thank you. Uh but it's the opposite
[13:05] of what these corporate institutions uh
[13:07] that Jason's talking about were at least
[13:10] attempting to do which is to be
[13:12] exceedingly careful with introducing
[13:15] information into the public domain.
[13:18] >> And and to expand on that and to answer
[13:20] your point, but something you mentioned
[13:21] earlier I think is crucial here which is
[13:22] like what are the incentives like what
[13:24] are we incentivized on a daily basis as
[13:26] journalists to do right? Um you know I
[13:28] work for a website that has no payw
[13:30] wall. uh we are primarily subsid you
[13:32] know our business models ads uh to
[13:35] paraphrase Mark Zuckerberg at his his
[13:37] famous hearing um and so you know we are
[13:40] incentivized to say what are people
[13:42] looking for that day what kind of
[13:44] information are they looking to seek um
[13:46] you know and that led us to make make
[13:48] decisions that you know sometimes I'm
[13:50] not in love with sometimes stuff that
[13:52] you probably don't even think has high
[13:54] value but you on that any particular day
[13:56] are looking for that information we are
[13:58] increasingly asking ourselves though Can
[14:01] chat GPT, can Google Gemini generate the
[14:04] article that we're writing? And if the
[14:06] answer is yes, then we don't write that
[14:07] article. That is very different than a
[14:09] decision I would have made even just a
[14:10] few years ago. Um to give a very
[14:12] definite example, um you know, HHS
[14:15] recently said it was cutting all this
[14:16] mRNA funding. We cover that news. We
[14:18] cover it sharp and quick because that's
[14:19] where we want to be. We want to be on
[14:20] the news, timely, relevant. Then we ask
[14:24] ourselves, what's the next story to
[14:25] write? Should we write a well, what is
[14:26] mRNA? That's a piece three years ago I
[14:28] absolutely would have assigned. People
[14:29] are going to be googling what's mRNA.
[14:32] It's not a piece I'm assigning anymore
[14:33] because that is what people I think are
[14:35] increasingly being going to be able to
[14:36] get from AI for better or worse. So
[14:39] that's time and resources that I'm
[14:41] dedicating to original reporting. So
[14:43] like I I kind of start liking the
[14:44] incentives. Suddenly I don't have to try
[14:46] to do information aggregation as much.
[14:49] We're focused on the news. We're focused
[14:50] on being timely and accurate. Then we're
[14:52] focused on finding original revoly
[14:54] impactful information. So if if there is
[14:57] a silver line to this and I think
[14:58] there's going to be rough road ahead in
[15:00] re in the coming years for media and for
[15:02] the businesses you know we run but if
[15:05] the incentives get realigned toward like
[15:06] highquality either timely or original
[15:08] journalism that that's a good thing.
[15:11] >> Jason you just inspired me with that and
[15:13] my question is is anyone having to
[15:15] change their writing to appeal to the AI
[15:17] engines? Is anyone having to like think
[15:20] okay if I write it like this if I do
[15:22] this claude or chat GBT is going to like
[15:24] it more?
[15:25] >> I'll jump in one more time. Uh uh your
[15:28] question prompts me to uh point to a
[15:31] thing that has emerged which is similar
[15:33] to the way that search engine
[15:34] optimization emerged with Google you
[15:36] know people trying to gain Google
[15:38] search. there's now essentially uh an
[15:40] industry forming around um you know
[15:43] trying to optimize uh
[15:44] >> generative yeah generative uh
[15:47] optimization geo
[15:49] >> G we'll call we'll call it GEO prompt
[15:51] jacking I prefer uh but you know the
[15:53] idea is if if you go and sprinkle out
[15:56] enough information maybe not all it's
[15:58] human readable it's white text it's you
[15:59] know burden code it's markdown somewhere
[16:01] on your site whatever uh if you pump
[16:04] enough of this stuff out there um then
[16:05] you can potentially change the outputs
[16:07] that come from Perplexity or Google etc
[16:10] who are indiscriminately scraping the
[16:11] internet right um and we're already
[16:14] seeing nation states doing this right
[16:16] we're already seeing uh brands do this
[16:18] we're already seeing uh in some cases
[16:20] people with political intentions doing
[16:22] this with the goal of essentially
[16:24] changing the outputs uh from those those
[16:27] mechanisms
[16:27] >> and that brings us to our next point
[16:30] being about the scale of misinformation
[16:34] AI allow for the creation of very
[16:36] realistic stuff that could be used by
[16:39] nation states or pressure groups uh to
[16:42] create misinformation. And I'm just
[16:43] wondering what your newsrooms are doing
[16:45] to protect your readers, to protect
[16:47] yourselves. And Nancy, I'll go with you
[16:49] first.
[16:49] >> No, it's I had an experience recently
[16:51] where I posted I think it was a story
[16:52] someone else had written about Zolan
[16:54] Mandani who's running for mayor in New
[16:56] York City and I got a paragraph l
[16:58] response on Twitter that said, "Oh no,
[17:00] you know, he you know, it was negative
[17:01] towards him." And I thought, "That
[17:02] doesn't sound like a human being would
[17:04] have written that." And I looked and
[17:05] it's a bot that simply responds to
[17:07] people who have, you know, I don't know
[17:10] if the Cuomo campaign is involved. I'd
[17:11] like to actually report that out. But um
[17:13] but that response you get back is just
[17:14] this, you know, a page of content
[17:16] telling you where you're wrong or making
[17:17] this particular point. You think like
[17:19] that does affect journalists, right?
[17:20] That you get this feedback being on
[17:22] social media, people are constantly
[17:23] yelling you. It does sort of affect how
[17:25] you do your day-to-day work. So I do
[17:26] worry that like, you know, we're we're
[17:28] writing, we're throwing it into the into
[17:29] the void and getting back just AI
[17:31] generated responses. At some point, you
[17:32] just say, "I don't want to do this job
[17:33] anymore.
[17:34] just not fun. So, I do worry about them.
[17:37] >> As a reporter, I do feel like it is I
[17:40] mean, always on me, but also on newsroom
[17:42] leadership to kind of set these
[17:43] standards. Like Nancy makes a great
[17:45] point. I mean, we do look at the
[17:46] internet to see how the public is
[17:48] reacting to things. And obviously, we do
[17:49] more reporting to confirm that, but it
[17:51] kind of just really hammers down how
[17:54] much you cannot rely on what you see
[17:55] online. It may look like tons of people
[17:57] are having this reaction somewhere, but
[17:59] it doesn't mean that an actual person is
[18:00] having that reaction now. So, I think
[18:02] just like I mean we're all always we
[18:04] should be this careful online. Like in
[18:05] the age of digital media, this has been
[18:07] the case. But it's really making the
[18:09] stakes higher yet again. I'm going to
[18:10] I'm going to play the the panel optimist
[18:12] though. I mean I don't want the internet
[18:15] to be flooded with garbage. Um but if it
[18:18] does become so then what you know we do
[18:21] becomes more important. And I think you
[18:23] know ideally people start realizing that
[18:25] and I think a lot of people now know
[18:27] they can't trust what they read on
[18:28] social media. They can't necessarily tr
[18:29] trust what they read on Google. Do they
[18:31] trust NBC News? We're working on that.
[18:33] Um, you know, I think, you know, media
[18:35] trust is not in a great shape and it's
[18:37] something that we think about and talk
[18:39] about every day. And, you know, again,
[18:41] to my previous point, I am hoping that
[18:43] as we all have to align ourselves to a a
[18:45] kind of like AI age that is going to
[18:47] mean like reestablishing like really
[18:49] clear value propositions of why are you
[18:51] coming to the New York Times, why are
[18:52] you coming to the Washington Post, why
[18:54] are you giving us or them money? Um, and
[18:57] and and that that will will be a good
[18:58] thing. But the downside of that is a
[19:00] very two-tiered information system in
[19:02] which the people who don't do that are
[19:04] very susceptible to exactly what Justin
[19:06] was talking about.
[19:06] >> So Jason, should lawmakers get involved?
[19:08] Should there be policy?
[19:12] >> That's that's tough. I I I don't think
[19:14] so. I I think you're encountering a lot
[19:15] of First Amendment issues if you start
[19:17] trying to to regulate what
[19:20] AI companies and what they generate
[19:22] should have to say or not say. Like to
[19:24] be honest with you, like I I that's not
[19:27] something I I know super deeply.
[19:29] The United States federal government is
[19:31] deeply involved right now, probably more
[19:33] involved than it ever has been in trying
[19:35] to manipulate the media industry and the
[19:38] tech industry. Uh this is the most
[19:40] sensorious uh you know administration
[19:44] we've seen perhaps in decades. um the
[19:46] types of attacks on independent media
[19:49] journalists um the the the ex you know
[19:53] essentially uh bribe like deals that are
[19:55] being extracted out of major media
[19:57] organizations um absolutely chilling so
[20:00] I think there is a policy um it may not
[20:03] be one that's you know bent towards uh
[20:06] ensuring that the truth is uh
[20:08] necessarily the priority of media
[20:09] organizations tech firms are also
[20:12] subject to the same uh type of jawbon
[20:14] that we're seeing right Now, um you're
[20:16] seeing is extraordinary pressure put on
[20:19] tech firms um types of uh scenarios that
[20:22] we never would have dreamed uh in terms
[20:24] of interference uh from the federal
[20:26] government. So the government's deeply
[20:28] involved in these questions um just
[20:30] perhaps not in the way that uh you might
[20:32] like.
[20:33] >> On the point of misinformation, I wonder
[20:35] I worry a little bit that it's we're
[20:37] comforting ourselves with the fact that
[20:39] um sometimes what AI produces I
[20:40] mentioned this is not high quality. I
[20:42] can't imagine that's always going to be
[20:43] the case, right? these firms are putting
[20:45] a lot of money into proving improving
[20:47] the quality of their results. So, do we
[20:49] like kind of, you know, delilute
[20:50] ourselves in saying, "Oh, we're always
[20:51] going to produce higher quality content
[20:52] that these AI tools um are going to
[20:55] publish and then miss, you know, sort of
[20:56] the the fact that at some point they
[20:58] catch up and uh and we're still, you
[21:00] know, we haven't sort of prepared
[21:01] ourselves."
[21:03] >> I think one thing that makes me hopeful
[21:05] about that is if you can build a brand,
[21:08] which I kind of hate saying that as a
[21:09] journalist. I don't view myself as a
[21:10] brand. I know that's kind of a thing now
[21:12] in our industry, but I do think if
[21:15] people feel attached to you, not
[21:17] necessarily because I mean, yes, you
[21:19] need to produce great content, but I
[21:20] don't know. I mean, a lot of people I
[21:22] think are New Yorker subscribers because
[21:24] they want to be New Yorker subscribers.
[21:25] And obviously, the New Yorker has a
[21:27] legacy that they're building on, but if
[21:29] you can just produce a product that
[21:31] people want to be the kind of person
[21:32] that reads that product, maybe that
[21:34] maybe that helps insulate you from that
[21:35] a little bit.
[21:36] >> I think the rise of Oops, am I out of
[21:38] out of juice? Oh, thank you. I think the
[21:39] rise of of news influencers, especially
[21:41] on Tik Tok, uh speaks to what you're
[21:43] talking about here is that I I do think
[21:45] people still like getting news from
[21:47] people, especially if they have a very
[21:48] clear idea of who they're getting it
[21:50] from. One of the things I wish we were
[21:51] doing better as a news organization, I
[21:53] think we're working toward is putting
[21:55] more of our reporters up front, like
[21:57] this is who you're getting your news
[21:58] from. Um you know, especially on the
[21:59] digital side, we don't do that. On the
[22:01] TV side, obviously, you know, the Lester
[22:03] Holtz of the world are very, very
[22:04] apparent. And those end up being people
[22:06] who like have really good trust scores,
[22:08] trust scores that exceed even like the
[22:10] the media organizations trust grow. I
[22:12] have a science fiction question to ask
[22:14] you all
[22:16] that wasn't in the notes.
[22:19] How bad is it going to get? You Nancy
[22:21] mentioned that you thought you know the
[22:24] layoffs would continue that there would
[22:26] be more consolidation media but how bad
[22:29] is it going to get before it gets better
[22:30] for journalism?
[22:33] Do we think that we're going to have
[22:35] another 80% reduction in media as NY's
[22:38] chart showed? Do we think we're only
[22:40] going to have about, you know, one major
[22:43] newspaper? How bad's it going to get?
[22:48] I mean, to NY's point, the the AI is
[22:51] just getting better. So, I can only I
[22:54] don't know how bad it's going to get,
[22:55] but I don't think that we've reached the
[22:57] bottom, certainly.
[23:00] >> And I think the trends aren't great
[23:01] because I think younger people have not
[23:02] developed a habit of paying for news um
[23:04] or seeking out news from particular
[23:06] sources. So, I don't think we're in good
[23:08] shape there.
[23:12] >> Um I obviously I am the pessimist and
[23:15] Jason, you're the optimist. And two of
[23:16] you seem, you know, right in the middle.
[23:18] I think that's good. Um, I'll I'll take
[23:21] a more pessimistic view. Um, artificial
[23:24] intelligence is, uh, great at
[23:26] concentrating power, right? Um, and
[23:29] right now we're seeing an extraordinary
[23:30] concentration of power into the hands of
[23:32] a handful of technology firms and we are
[23:35] seeing the combination of that
[23:36] concentration of power with state power,
[23:39] right? Um so if that trend continues I
[23:42] think it's very bad news uh for you know
[23:46] the information independent information
[23:48] ecosystem generally um and I suspect we
[23:51] will also see more attacks on free and
[23:53] independent media in this country um you
[23:56] know uh right now we're seeing a kind of
[23:58] mini occupation of the nation's capital
[24:00] apparently uh the Pentagon has drawn up
[24:03] plans to send rapid response forces to
[24:06] uh other cities in the country um you Uh
[24:09] there's a headline on the Wall Street
[24:10] Journal about uh uh you know the the the
[24:14] Trump administration uh kind of moving
[24:17] in a direction of of sort of state
[24:19] capitalism particularly with regard to
[24:21] uh the relationship to tech firms. Um
[24:24] the context is changing quickly. Um and
[24:26] I think the news and information piece
[24:28] is just part of that.
[24:29] >> So I think we could call it a panel.
[24:31] Thank you all for coming.
[24:33] [Applause]
[24:33] [Music]
