John Herrman
LightNews LightNews
jwherrman.bsky.social
John Herrman
@jwherrman.bsky.social
8K followers 560 following 620 posts
posting about posts at new york magazine. have me on your podcast!
Posts Replies Media Videos
jwherrman.bsky.social
guy with no tools who doesn’t know how many cylinders his car has after ten minutes of ChatGPT: I can fix this

guy’s girlfriend after consulting her ChatGPT with memory mode enabled: This is exactly like when we went camping and there’s actually a term for this behavior
November 7, 2025 at 10:42 PM
jwherrman.bsky.social
AI companies want to mediate the entire economy. Their competitors aren't happy about this nymag.com/intelligence...
This week, Amazon answered clearly in the affirmative, in the form of a lawsuit against Perplexity. The language is spicy: Amazon brings this action to stop Perplexity AI, Inc's ("Perplexity" or "Defendant" persistent, covert, and unauthorized access into Amazon's protected computer systems in violation of federal and California computer fraud and abuse statutes. This case is not about stifling innovation; it is about unauthorized access and trespass. It is about a company that, after repeated notice, chose to disguise an automated "agentic" browser as a human user, to evade Amazon's technological barriers, and to access private customer accounts without Amazon's permission. Fraud! Abuse! Trespass! Evasion! This is a threatening lawsuit — it later refers to Perplexity as an "intruder" — but it's also a useful window into Amazon's thinking on AI and where its business fits into a theoretical world where more people do more things with chatbots and agents. It might not be right to say the company is worried, exactly. But it's certainly clear about what it doesn't want to happen, which happens to be precisely what some AI companies clearly want. Most broadly, AI companies are hoping to end up in a situation where their products are the default interface for all sorts of things, and browser agents represent a small but aggressive move in that direction. A world where all purchases flow entirely through another company's interface, chatbot or otherwise, is something that Amazon would prefer to avoid, or at least have a say in, which is perhaps why the company's lawyers are talking about Perplexity as if it's built an automated ticket-scalping app, a sneaker-sniping bot, a data scraper, a data-exfiltration tool, or rip-off interface for its product catalogue. Earlier this week, Perplexity argued on its site: Amazon wants to block you from using your own AI assistant to shop on their platform ... Amazon should love this. Easier shopping means more transactions and happier customers. But Amazon doesn't care. They're more interested in serving you ads, sponsored results, and influencing your purchasing decisions with upsells and confusing offers. Suggesting that "Amazon should love this" and then describing the ways that it might cause them to make less money is sort of funny, but between the lawsuit and Perplexity's response we can get a pretty clear sense of what's going on here: an attempt by Amazon to stop a marginal player from setting a precedent it doesn't want, and a clear signal to bigger ones — OpenAI, Google, and other firms Amazon already competes and partners with - that, if they want their tools to touch Amazon, they'll have to do it on Amazon's terms. This is a sort of conflict we can expect to see a lot of, and soon.
November 7, 2025 at 3:34 PM
jwherrman.bsky.social
So many bizarre AI products are like this but it's really important to understand why: Companies like OpenAI want the world to give in to their total success before they've actually achieved it nymag.com/intelligence...
As I spent time watching ChatGPT bonk its way through various web interfaces, I also found myself thinking of self-driving cars. A browser that pretends to be a person at the input level - moving a cursor, scrolling a human GUI - felt less like a Waymo, in which an unattended steering wheel turns as a result of actions taken by systems closer to the road, than a regular car with a humanoid robot sitting in the driver's seat. Again, it's pretty interesting to watch! But it also makes you wonder: Why are we doing this that way? Aren't there better ways for machines to talk to each other? The answer, Miller suggests, comes down to the "incentives of their makers more than the intrinsic value of the technology." (The Browser Company was recently acquired by software-maker Atlassian.) For OpenAI, building systems that can execute complex commands on behalf of users is the whole ball game - it's the path to wide-ranging automation and/or AGI, depending on which definition of the term the company is going with that day. An AI model that can provide useful information in a chat window, or handle tasks clearly outlined by the user, is a useful product, but the prospect of an Al model that can productively and proactively interact with the world around it in ways comparable to a human - or, more specifically, an employee — is where trillion-dollar valuations come from. To get where it wants to go, though, OpenAl has a number of challenges. Some are widely discussed and frustratingly hard to pin down, revolving around benchmarks, varying definitions of model capability, and predictions about scaling. Others are more banal: To answer questions more usefully, for example, chatbots tend to do better if they have more data about users; likewise, to execute tasks on their behalf, they need to operate in an environment where users are logged in to the various services they use to work and live their lives. They need a breathtaking amount of access and permission, in other words. ChatGPT isolated in a chat window doesn't have that, and it takes a long time to draw out of users, if they're willing to offer it at all. ChatGPT as a default browser - authenticated in dozens of different sites, payment methods at the ready, or perhaps even logged into a work environment - does. (Such agents also create, as many in the AI space have pointed out, a potential security nightmare.)
October 30, 2025 at 3:13 PM
jwherrman.bsky.social
The success of Ring is just an incredibly American story: Individualism expressed through consumer habits resulting in the voluntary creation of a national surveillance network. Add home security cameras to the list with guns, cars
In Ring's case, customers largely saw themselves as on the same side as Amazon in a fight against porch piracy and for convenience, if they thought of their purchase in such terms at all. The privacy-based case against buying a Ring camera wasn't terribly persuasive from the perspective of potential customers, because they were the ones doing the surveillance, from their property, protecting their stuff and homes. Ring cameras invited users to adopt an Amazonian logic of their own, in which total stoop awareness is, clearly, a good and desirable thing. Barely a decade after Ring's doorbell-camera pitch first aired on Shark Tank, though, Amazon and millions of its customers have haphazardly and perhaps not entirely consciously teamed up to build something strange, unprecedented, and very Amazon: a crowd-sourced, nationwide, neighborhood surveillance network to which the government now has a set of keys.
October 27, 2025 at 6:21 PM
jwherrman.bsky.social
Interesting how basically every Amazon project — true of most tech neo-conglomerates, to be fair — comes back around to surveillance nymag.com/intelligence...
In various nearby industries, Amazon is regarded as a leader in employee surveillance and "algorithmic management" and is likewise held up as an example of its possible effects on workers (more productive, less content, and less likely to organize). In a brief rendering demonstrating how its delivery glasses will work, Amazon shows how worker surveillance will soon extend outside of the vehicle, right up to the doorstep. Which brings us to the second bit of news, from TechCrunch: Amazon's surveillance camera maker Ring announced a partnership on Thursday with Flock, a maker of AI-powered surveillance cameras that share footage with law enforcement. Now agencies that use Flock can request that Ring doorbell users share footage to help with "evidence collection and investigative work." Ring cameras are an underrated Amazon success story: With an appealing pitch — see who or what is on your doorstep, even if you're not home — the company sold millions of units and constructed a massive surveillance network with numerous benefits for the company itself. They're a way to counteract package theft and to make it easier for customers to receive deliveries. They also provide an additional way for the company and its customers to surveil its workers, meaning that Amazon's glasses aren't just extending its "end-to-end" delivery apparatus — they're closing the loop in its employee-monitoring ecosystem. Ring cameras were, for obvious reasons, always interesting to law-enforcement agencies, with which Amazon has had a generally cooperative but heavily scrutinized and limited relationship, at least until recently. Now, by partnering with companies like Axon and Flock, which operate nationwide fleets of license-plate scanners and work with local, state, and federal law-enforcement agencies, including ICE, the company is making its surveillance network widely and comprehensively available to the state.
October 27, 2025 at 6:21 PM
jwherrman.bsky.social
if you want a picture of the future, imagine your own face staring back at you from the chumbox, forever nymag.com/intelligence...
Virtual try-ons are actually showing up all over the place in retail — you'll find them on both luxury-brand apps and Shein — and they tend to suffer from the same issues, given that they're essentially mashing images together: The clothes don't fit quite right; the lighting is always strange; generations can take a little while; body types are approximated but rarely nailed. But they might be good enough to help rule something out or consider something new, and they're certainly good enough to be compelling and occasionally fun to use. The temptation to merge virtual try-ons with ads — to follow users around the internet not just with evidence that you know a lot about them but with actual photos of themselves using new products — is surely massive but is currently constrained by two things. First, generating such images or videos is computationally and literally expensive. And second, putting users' faces in, say, a branded Instagram Reel would probably feel to many like a violation. (For now, Google's explainer about its virtual try-on feature marks a boundary far more conservative than that, emphasizing that the feature won't currently show up on sponsored products.) 凸 One of those constraints will probably go away as AI models become more efficient. The other is a matter of expectations and norms around privacy, which, if the last two decades are any guide, have a tendency to either erode or get bulldozed. Given the hunger for return on AI investment and the close relationship between the advertising industry and AI - Google and Meta, for example, are advertising and AI giants, and it's hard to imagine the latter holding out too long on a new technology that might improve ad performance — it's probably a matter of time before someone breaks the seal and starts using customers to market to themselves at scale. Small companies are already trying something weird and uncomfortable. The bigger ones probably aren't far behind.
October 16, 2025 at 3:49 PM
jwherrman.bsky.social
people don't like being told what they can't post, but they REALLY don't like being told what they can't generate nymag.com/intelligence...
a collage of negative reviews of Sora focused on its "guardrails"
October 14, 2025 at 2:40 PM
jwherrman.bsky.social
implied agentic ai pitch: we will lure users with unrealistic promises, and in the process turn them into spammers that destroy the entire existing marketplace. then... profit?
The bet by investors like Y Combinator, in other words, is on a broadly familiar form of disruption but with a couple of twists. Generative AI may expose some incumbent firms to outpricing, outselling, and outmaneuvering by start-ups that embrace it, while more and more people acclimate to strange, new, chatcentric modes of computing. In the meantime, it may also help glut, jam, and temporarily destroy the entire markets in which they're operating. It's a version of creative destruction in which the destruction is preemptive, indiscriminate, and maybe looks a bit like sabotage — with a little bit of help from all of us, of course.
October 10, 2025 at 1:13 PM
jwherrman.bsky.social
trying to think through what's unusual about so many AI startups, and about the gaps between their pitches, the real world, and what they would actually need to succeed nymag.com/intelligence...
This attitude toward externalities — not my problem, and in any case worth it in exchange for a small advantage — follows the approximate logic of a spammer and often comes wrapped in the language of AI hustle culture. It's also understandable from the perspective of a job seeker who feels constantly thwarted by automated systems employers use that seem to treat seekers with similar indifference or contempt, or by platforms like LinkedIn and Indeed that, while nominally intended to connect two parties with shared interests (one needing specific services, the other offering them), can feel more like social-media-style black holes for engagement. It's an escalation that will likely be met with more escalation: countermeasures by job-listings platforms and hirers to prevent access by AI agents; more aggressive automated filtering; different hiring routines altogether, making it even harder to get through the door to a coveted interview. Mercenary (and slightly deceptive) automation tools like this, which are being pitched all over right now and already wreaking havoc in, for example, online dating, depend on two temporary circumstances to work, if they ever actually do: (1) that most other people don't have access to them, giving the user an edge and (2) that the people and parties on which they're used will tolerate and take no action against them. In other words, if you take their pitches at face value, they're pretty obviously doomed in the medium term, in the sense that they'll either be rejected by the systems they operate in or simply ruin them for everyone. Taking stock of the first few years of mainstream AI deployment, though, raises an important question. What if that's sort of the point? Or at least a world worth thinking about in a more thorough, long-term way? Generative image and video tools, for example, have significantly degraded social-media platforms, allowing bad actors and regular people to fill them with slop, intensifying existing problems with spam and deceptive content while thwarting old solutions. And, hey, look at that: Suddenly, OpenAI and Meta are launching new social networks based on AI, on which posting generated content is the point, not a problem to be solved. Generative AI may be placing immense stress on educational institutions and worsening the already strained relationships between teachers and students, but wait — every AI company is selling ed tech now.
October 10, 2025 at 1:13 PM
jwherrman.bsky.social
it's only been a week but I'd bet a lot of early users are having experiences like this with Sora: manic, exploratory onboarding followed by something like a hangover. it's a pattern! nymag.com/intelligence...
Sora is presumably extremely expensive to run, hence OpenAl's use of a chained invite program to roll it out. In its early form, it brings to mind the early days of image generators like Midjourney, the video model for which Meta is now borrowing for Vibes. Like Sora, Midjourney in 2022, was a fascinating demo that was, for a few days, really fun to mess with for a lot of the same reasons: A vast majority of the images I've generated have been jokes — most for friends, others between me and the bot. It's fun, for a while, to interrupt a chat about which mousetrap to buy by asking a supercomputer for a horrific rendering of a man stuck in a bed of glue or to respond to a shared Zillow link with a rendering of a "McMansion Pyramid of Giza..? ...I still use Midjourney this way, but the novelty has worn off, in no small part because the renderings have just gotten better — less "strange and beautiful" than "competent and plausible." The bit has also gotten stale, and I've mapped the narrow boundaries of my artistic imagination. Playing with Sora is a similar experience: a destabilizing encounter with a strange and uncomfortable technology that will soon become ubiquitous but also rapidly and surprisingly banal. It also produces similar results: a bunch of generations that are interesting to you and your friends but look like slop to anyone else. The abundant glitches, like my avatar's tendency to include counting in all dialogue, help make the generations interesting. Many of the videos that are compelling beyond the context of their creation are interesting largely as specimens or artifacts — that is, as examples of how a prompt ("sam altman mounted to the wall like a big mouth Billy bass, full body") gets translated into ... something. (As one friend noted, many of these videos become unwatchable if you can't see what the prompt was.)
October 6, 2025 at 4:33 PM
jwherrman.bsky.social
OpenAI's maximalist "asking for infinite money" strategy has a legal counterpart, too nymag.com/intelligence...
The opt-out process for the new version of Sora means that movie studios and other intellectual property owners would have to explicitly ask OpenAI not to include their copyright material in videos the tool creates. This is pretty close to the maximum possible bid OpenAI can make here, in terms of its relationship to copyright — a world in which rights holders must opt out of inclusion in OpenAI's model is one in which OpenAl is all but asking to opt out of copyright as a concept. To arrive at such a proposal also seems to take for granted that a slew of extremely contentious legal and regulatory questions will be settled in OpenAI's favor, particularly around the concept of "fair use." Al firms are arguing in court — and via lobbyists, who are pointing to national-security concerns and the Al race with China - that they should be permitted not just to train on copyrighted data but to reproduce similar and competitive outputs. By default, according to this report, OpenAI's future models will be able to produce images of a character like Nintendo's Mario unless Nintendo takes action to opt out. Questions one might think would precede such a conversation — how did OpenAI's model know about Mario in the first place? What sorts of media did it scrape and train on? — are here considered resolved or irrelevant. Coming from Altman, though, it assumes an additional meaning: He would very much prefer that his company not be liable for potentially risky or damaging conversations that its software has with users. In other words, he'd like to operate a product that dispenses medical and legal advice while assuming as little liability for its outputs, or its users' inputs, as possible — a mass-market product with the legal protections of a doctor, therapist, or lawyer but with as little responsibility as possible. There are genuinely interesting issues to work out here. But against the backdrop of numerous reports and lawsuits accusing chatbot makers of goading users into self-harm or triggering psychosis, it's not hard to imagine why getting blanket protections might feel rather urgent right now. On both copyright and privacy, his vision is maximalist: not just total freedom for his company to operate as it pleases, but additional regulatory protections for it as well. It's also probably aspirational — we don't get to a copyright free-for-all without a lot of big fights, and a chatbot version of attorney-client privilege is the sort of thing that will likely arrive with a lot of qualifications and caveats. Still, each bid is characteristic of the industry and the moment it's in. So long as they're building something, they believe they might as well ask for everything.
October 1, 2025 at 1:27 PM
jwherrman.bsky.social
strong agree, from 2017: opaque commercial social platforms becoming the most vivid, accessible representation of discourse, democracy, the public sphere, the media, etc was a massive accelerant for cynicism/resentment/despair www.nytimes.com/2017/08/21/m...
September 18, 2025 at 2:34 PM
jwherrman.bsky.social
The lesson isn't "EVERYONE is being radicalized by the internet" — a similarly appealing, pat, useless framework except for assigning blame and justifying censorship and privacy crackdowns — but rather a warning against imagining the internet as a distinct place or force. We've got bigger problems!
Of course, we don't really expect the many thousands of people saying this kind of stuff about the killing of a widely beloved and despised famous person to one day resort to violence. We understand them as downstream not just from a specific event and its fallout but from a terrifying and heavily politicized decline in Americans' trust and regard for one another. (Also, it should be said, as relatively rare public expressions of private sentiments more people might confide to friends or family, who are more likely to agree, understand, or simply feel comfortable knowing that they don't mean much.) This should give us pause about the ritual of assigning such posts so much value in cases where violence actually occurs. As a general theory explaining the ills of the world, "online radicalization" is a comforting fiction without predictive power, a way to contain shocking events with complicated, overlapping motivations inside arbitrary digital boundaries, a tempting method for dismissing deeper and more sinister problems as mere symptoms of a vague and external online menace, and as a tool for political blame or vindication. "The internet" can usually no more provide us with the full story of why someone decided to kill another person than their isolated experiences at work, school, or home, which we hardly struggle to see as entwined and inseparable. But the grim incentives and cynical priorities of the commercial platforms that make it up might help explain how we come to see one another as irredeemable.
September 18, 2025 at 1:16 PM
jwherrman.bsky.social
Fascinating to watch people strain to assemble a "radicalized online" theory from thin and incomplete evidence as they post amid some of the most openly vengeful, bloodthirsty, "radical" social feeds in memory nymag.com/intelligence...
Was Tyler Robinson radicalized online? To whatever extent it's a meaningful, coherent question, the best answer we have, based on what little evidence we have — and with the expectation that more will be revealed in the coming days, weeks, and months — is: maybe? We don't know. If and until we find out, though, the days since Kirk's shooting have posed a related question: What about everyone else? If the popular theory of online radicalization is that people lose touch with their humanity - or are merely driven to ideological extremity or total alienation - in insular, ideological accelerative communities, there were abundant examples of this phenomenon in the week following Kirk's shooting. Seconds after Kirk was killed, a TikTok influencer streamed the scene mere feet away from the stage where it happened, frantically commanding viewers to subscribe to his channel as panicked bystanders screamed and fled. In the early stages of the manhunt, politically motivated users on X, and on forums like 4chan and Kiwi Farms, began constructing theories about how the shooter must have been trans and therefore predisposed to violence, laundering a preexisting shared obsession into the broader social-media conversation. Elsewhere on social media, defensive leftists convinced one another, with scant evidence and clear motivation, that Kirk must have been killed by a follower of far-right personality Nick Fuentes, who had feuded with him for years. (It seems like a version of this theory may have found its way to the writer's room at Jimmy Kimmel Live!, drawing the attention of FCC chair Brendan Carr and leading ABC to take the show off the air "indefinitely.") Before and after Robinson was identified, a small but not insignificant number of people cheered Kirk's death on public social-media channels, while right-wing internet users responded with an unprecedented mass doxxing campaign, collecting these posts and trying to get the people behind them fired, kicked out of school, or deported. This was all cheered on by the J.D. Vance - "call them out, and hell, call their employer," the vice-president said while hosting Charlie Kirk's show. It was gleefully justified as revenge for the left's years of"cancel culture," and was accordingly broad, sweeping up people who had merely criticized Kirk, as well as bystanders who had, in fact, said nothing at all. Elsewhere on social media, and especially on X, some users simply — and comfortably — called for direct retribution and declared that the time had come for a national divorce and civil war. Animated, horrified, inspired, or merely unmoored into a state of disorientation by the shocking (and widely viewed) murder of a polarizing figure, a vast range of people revealed, on purpose or by accident, how alien to one another their normative experiences had become, online and off. Many surely wondered, from each edge of this nightmarish discourse — although perhaps not as much as people who didn't enter the fray at all — in what world are people posting this stuff? In death as in life, Charlie Kirk has left a lot of Americans struggling to understand how their fellow citizens could possibly say or believe the things they do. I don’t mean to suggest equivalence here. An obsessive, metrics-driven influencer is a freakish cautionary tale, not a militant slipping into terroristic threats. You might question the character or judgment of someone who publicly celebrates the death of public figure they hate, or vents crudely or angrily about their political opponents after seeing a video of the murder of an online personality they love or identify with, but they’re not participating in an organized online campaign, supported by elected officials, to have their enemies “thrown from civil society,” kicked out of school, fired from their jobs, or stripped of their driver’s licenses. Fearful elected Democrats have been far more careful than scattered liberals and leftists on social media, retreating to platitudes about how violence is bad and speech should never be met with violence, and have been harshly censured by Republicans in cases where they offered anything but praise for Kirk, as in the case of representative Ilhan Omar — the right is, as usual, more aligned. The president claimed that the “radical left” was “directly responsible” for an act of “terrorism” and pledged to go after people and organizations that “contributed to this atrocity and to other political violence.” His closest adviser, Stephen Miller, was more direct and spoke in the same voice as the burn-it-all-down contingent of Kirk’s online fandom. Kirk’s death was the result of “a vast domestic-terror movement,” he told Vance on Kirk’s show, before promising to use “every resource we have at the Department of Justice, Homeland Security, and throughout this government to identify, disrupt, dismantle, and destroy these networks and make America safe again for the American people.”
September 18, 2025 at 1:16 PM
jwherrman.bsky.social
OpenAI's "what is ChatGPT, anyway?" study is a little weird, sort of funny, but actually illuminating nymag.com/intelligence...
There are a few major caveats here. OpenAI's own researchers worked on the paper with Harvard economist David Deming under the auspices of the National Bureau of Economic Research (in other words, the company is comfortable with this paper's findings). Additionally, the research, which is mostly an attempt to classify and sort a large set of messages, was done substantially by OpenAl's own models, which both automated the process and, the researchers say, helped preserve user anonymity. "No one looked at the content of messages while conducting analysis for this paper," the researchers claim, although they did validate their automations against human analysis used in smaller, previous studies we've talked about here before. The researchers' methodology is both fascinating and could insufficiently but not entirely inaccurately be characterized as: "We asked ChatT The picture that emerges from this data matches this thesis pretty closely: ChatGPT, for many of its users, is a way to access, remix, summarize, retrieve, and sometimes reproduce information and ideas that already exist in the world; in other words, they use this one tool much in the way that they previously engaged with the entire web - arguably the last great "cultural and social technology" - and through a similar routine of constant requests, consultations, and diversions. One doesn't get the feeling from this research that we're careening toward uncontrollable superintelligence, or even imminent invasion of the workforce by agentic Al bots, but it does suggest users are more than comfortable replacing and extending many of their current online interactions - searching, browsing, and consulting with the ideas of others - with an ingratiating chatbot simulation.
September 16, 2025 at 2:36 PM
jwherrman.bsky.social
clear act of war against old people and their families www.theverge.com/news/769460/...
September 4, 2025 at 2:04 PM
jwherrman.bsky.social
new chatbot? call it an agent. new software feature? sure, yes, agent. Old feature that's been there for years? It can be an agent if it wants to be, and how dare you suggest otherwise nymag.com/intelligence...
A few of Grammarly's new agents might fit such a definition, if you squint a little. The "AI Grader" tool claims to be able to estimate an assignment's grade by "looking up your instructor, "reviewing public teaching info," and "identifying key grading priorities." Setting aside whether such a feature is ethical, useful, or even functional - Jane Rosenzweig, director of the Harvard College Writing Center, found nothing redeeming in testing - such a feature might fit a bare description of an agent. You give it a task and it does a couple of unpredictable things in its attempt to fulfill it. Same for the "Citation Finder, a tool that will, after your work is done, backfill it with "legit sources to back your claims." (Now is as good a time as any to ask: What the hell are we even doing with this stuff?) A Grammarly feature that claims to offer "feedback inspired by real experts" seems to drift further from "agency." Then there are the rest of the new agents: simulated "reader reactions," a "humanizer," a "paraphraser." These sound an awful lot like features already contained in Grammarly, as well as things you could get by pasting your text into any major chatbot and asking for it. (Are the new included summarization features in iOS, Android, Windows, Office, and Google Docs agentic too?) Finally, Grammarly suggests, there's a "proofreader" agent, which can "refine grammar, clarity, structure, and more as you write." It's hard to find the agent here or even something new. Proofreading has been the entire thing with Grammarly for years, well before the company was talking about generative Al at all.
August 22, 2025 at 3:42 PM
jwherrman.bsky.social
As reported on @nymag.com earlier this month! nymag.com/intelligence... (no beef, @theinformation.com has real confirmation and lots of interesting details)
August 22, 2025 at 2:53 AM
jwherrman.bsky.social
I'm insufferably, interminably on-record arguing that AGI discourse is alienating, historically untethered, and misleading (nymag.com/intelligence...). But now of its biggest boosters — and, as @himself.bsky.social points out, an influential figure in AI *and* foreign policy — says: ok, yeah
As an insider effort to regulate the frequently wild messaging coming from the Al industry, this is surprising rhetoric. It is consistent with (and actually cites) critical work arguing that while recent developments in Al represent a major technology with great disruptive potential, current models might not be on the cusp of rendering the entire economy obsolete and/or wiping out or subjugating humanity once machines have established themselves as a superior species. The argument that stories about AGI are alienating, distracting, and disempowering - no objection here! - is likewise sort of wild to hear coming from Schmidt, who just this year warned that Al models are beginning "recursive self-improvement" and are already "learning how to plan, and they don't have to listen to us anymore." But political scientist and Al theorist Henry Farrell makes a case for the deeper significance of such a pivot from Schmidt, who last year co-authored a book with Henry Kissinger about Al, and who Farrell has argued could be "the most influential American foreign policy thinker of the early 21st century. In short, Schmidt was instrumental in forming the "mind meld" between Silicon Valley and national-security policymakers, which held that self-improving AI, and AGI, represented a fast-approaching "inflection point" in a race between the U.S. and China, the winner of which would gain a "technological advantage that would secure its long-term dominance." But, he writes, "if the AGI bet is a bad one, then much of the rationale for this consensus falls apart. And that is the conclusion that Eric Schmidt seems to be coming to."
August 20, 2025 at 5:55 PM
jwherrman.bsky.social
Altman says all sorts of things, all the time. But the hard pivot away from AGI/superintelligence rhetoric by Eric Schmidt is genuinely jarring, sort of funny, and I think telling nymag.com/intelligence...
In the last week, two prominent voices in Al attempted to make some adjustments to the AI discourse. Asked if "investors as a whole are overexcited about Al," OpenAI's Sam Altman answered "yes" to a group of gathered reporters, according to the Verge. Then he used that word. "When bubbles happen, smart people get overexcited about a kernel of truth," he said. "If you look at most of the bubbles in history, like the tech bubble, there was a real thing. Tech was really important. The internet was a really big deal. People got overexcited." It would be a mistake to read this as a major departure from Altman's wide-ranging and often inconsistent public musings about Al: His intention here, after a rocky rollout for his company's latest model, GPT-5, seems to be to situate OpenAI outside of a possible bubble, implying that his company has more in common with dot-com Amazon than with Pets.com or Worldcom. Still, it's significant that Altman broached the subject, and there are signs that a generally bullish Wall Street has taken notice: In a story about so-called "disaster puts," Bloomberg reports that "options traders are increasingly nervous about a plunge in technology stocks in the coming weeks and are grabbing insurance to protect themselves from a wipeout." In another notable shift, former Google CEO Eric Schmidt, who has in recent years become an influential and provocatively severe figure in the Al world, co-authored a piece for the New York Times with tech analyst Selina Xu arguing that Silicon Valley "needs to stop obsessing over superhuman Al." Just a few months ago, in another op-ed, Schmidt argued that "even without a consensus about a precise definition, the contours of an AGI future are beginning to take shape," suggesting that "artificial general intelligence" could "usher in a new Renaissance." This week, he and Xu urged their peers to rethink such messaging and suggested that narratives like this actually gave them "pause": It is uncertain how soon artificial general intelligence can be achieved. We worry that Silicon Valley has grown so enamored with accomplishing this goal that it's alienating the general public and, worse, bypassing crucial opportunities to use the technology that already exists. In being solely fixated on this objective, our nation risks falling behind China, which is far less concerned with creating A.I. powerful enough to surpass humans and much more focused on using the technology we have now.
August 20, 2025 at 5:55 PM
jwherrman.bsky.social
www.theguardian.com/uk-news/2016...
Google has thanked an 86-year-old British woman who proved old-fashioned manners have a place in the modern world when she typed "please" and "thank you" in an internet search. May Ashworth's grandson Ben John found her laptop open and took a photo of the unusually polite online request. She was asking for a translation of the Roman numerals MCMXCVIII.
August 20, 2025 at 1:11 AM
jwherrman.bsky.social
just checking in with one of the most influential philosophers alive today
August 19, 2025 at 2:46 PM
jwherrman.bsky.social
A little pushback on the "AI researchers are like NBA stars" stuff: Meta, which is driving this narrative, is uniquely toxic. In an industry that contains Elon Musk! nymag.com/intelligence...
It also hasn't helped that Zuckerberg's recent, Altman-esque attempts at Al prophesying are even less convincing from within the industry than from outside it, and to the extent he has tried to articulate a vision - replacing friends with Al, a vaguely defined "superintelligence" that will somehow mainly benefit individuals - it comes across as borrowed and incoherent even by the low standards of CEO AI prognostication. And whatever motives you might ascribe to the leaders of Al firms, the talent they're pursuing tends to be — by a few definitions of the word - somewhat idealistic. If they're worried about Al risk, they're probably more inclined to work for someone like Anthropic's Dario Amodei than Mark Zuckerberg circa 2025. If they're interested in pure research, they're probably going to want to work for people they see as fellow pure researchers - people like Google's Demis Hassabis or a founder like Ilya Sutskever. If they're a bit more accelerationist and spend a lot of time absorbing the sort of Al discourse that's popular on X - or if they're just looking at where concentrations of like-minded talent already exist - they might be more inclined to work at OpenAI. Gruesome as parts of it may be, even XAI has a coherent pitch to a certain sort of "cracked" Al engineer: lots of resources, few guardrails, proximity to Elon Musk, and a corresponding ideological mission. Perhaps most importantly, setting aside various specific motivations and theories of what the Al arms race really is, you won't find a lot of people in the industry who would bet on Meta winning it, at least right now. (Though one could argue that the company's entertainment-centric, slop-and-advertising-adjacent approach to actually monetizing Al so far represents a realistic if disheartening path toward extracting value from LLMs.) That Mark Zuckerberg has a weaker future-of-Al pitch than the MechaHitler guy goes a long way toward explaining his company's outlier offers (recent stories about internal materials allowing "sensual" chats with kids certainly can't help). But elements of Meta's reputation that are less visible outside of the tech industry are playing a major role as well. At Forbes, Rashi Shrivastava and Richard Nieva describe deeper problems, reporting on constant internal restructurings (the fourth in six months is underway, according to the Information), unclear priorities, and bruising demands to work on products that, so far, have been poorly received. While punishing "9-9-6"-style work schedules are becoming more common at AI firms, employees are often motivated by genuine enthusiasm, fear of their companies losing a high-stakes race, and, of course, lots of equity tied to their work's success. Since 2023, Meta's general orientation toward its employees has become much harsher, with layoffs and public warnings from leadership sharply diverging from the company's financial performance. A widely circulated internal memo from a (now former) Al researcher at Meta warned that a "metastatic cancer" was threatening the company's work culture. Last week, in a post meant to push back against "sensationalist" reporting on the memo, he described his intention to draw attention to the "unhealthy" traits of the Al organization and wasn't actually much gentler: • The fear that people feel on a daily basis of reviews and getting fired, commenting on a lack of safeness people felt, which is crucial for morale. • The necessity of a culture and processes that enable 'big projects' to come to fruition. • The management culture not promoting a sense of camaraderie among employees, leading to a lack of sense of belonging. • Instability in team assignments, leading to experience not building up and crystallizing over time.
August 19, 2025 at 2:17 PM
jwherrman.bsky.social
The administration will have brokered this deal, and other companies have already bent to its will for far less. Why wouldn’t the Trump administration do with TikTok what the ban accused China of doing first?
The sale is (as far as we know) still being shepherded by Vance, who says he's pursuing a deal that would be "protective of Americans' data privacy and America's national security," and who earlier this year said Big Tech companies were "on notice" over censorship and AI models should be "free from ideological bias." (Recycling old comments by his mentor and patron Peter Thiel, he has also suggested that Al is a "left-wing" technology, never mind the recent "woke AI" executive order.) The companies making the bid, which would leave ByteDance with a minority stake in TikTok, are led by three outspoken Trump supporters and MAGA donors: Oracle's Larry Ellison, Blackstone's Stephen Schwarzman, and Andreessen Horowitz's named partners. Earlier this year, Meta settled a lawsuit with Trump over his bans from Facebook and Instagrar with a sizable donation to his presidential library. Just last week, Meta settled an AI defamation lawsuit with pro-Trump influencer Robby Starbuck, who will now advise the company on its efforts to "mitigate ideological and political bias." Elon Musk agreed to a settlement on behalf of Twitter's previous ownership, too, and Google is widely anticipated to be up next.
August 13, 2025 at 1:44 PM
jwherrman.bsky.social
I’m not the first person to suggest this but I’m worried it’ll take people by surprise: What happens if TikTok goes full MAGA? nymag.com/intelligence...
The transfer of control of TikTok to a group of Trump-aligned American firms in the name of national security would be a more severe outcome than a settled presidential lawsuit or regulatory harassment, and beyond the natural alignment between owners and the government, it could leave the administration with unprecedented leverage over the new firm. Vance would have brokered the deal in part, of course. But there's also a good chance that the new firm, at least as outlined in press reports so far, wouldn't be fully compliant with the letter of the TikTok ban as written. Instead, it would merely be declared compliant by the Department of Justice, which could — as is the case with the agency's deferrals of the ban — simply change its mind. A new TikTok owned by a syndicate of Trump allies probably wouldn't need to be told what sorts of things the administration does or doesn't want to see on one of the most popular social media apps in the world, of course. But they'll be in charge of a powerful entity that the president could in some way see as his - and which he already believes helped him win the election. There's other recent context to consider, too. Musk spent billions of dollars to make X a more agreeable platform for his worldview and Trump's; Mark Zuckerberg, for reasons of his own, overhauled Meta's moderation systems after Trump's reelection. We don't yet know what sort of platform awaits users of neo-TikTok or M2, or even if the Chinese government will allow such an arrangement to exist. If it does, though, the new entity will have every reason - and perhaps a natural inclination - to be sensitive to the president's desires and grievances. More to the point, it might not have the option to ignore them. The rationale for the TikTok ban, as articulated in Trump's initial 2020 executive order, was that it would censor "politically sensitive" content and could be used "for disinformation campaigns" under the direction of ByteDance's home government. (It's also an article of faith in the MAGA movement that American tech companies already conspired with the Biden administration in a comparable way.) At this point, shouldn't we expect the Trump administration to exercise its available powers to the fullest, and then some?
August 13, 2025 at 1:41 PM