 |
|
|
|||||
Hello again, and thanks for reading Fast Companys Plugged In. When you think about it, training AI to use the web might be the single most impactful way to expand its power. So much of what we do todayfrom buying products of all kinds to managing every aspect of our personal datawe do online. If a piece of software could handle that work at least as well as a human, it could be a far more essential assistant than any existing AI tool. Web savvy is key to the tech industrys current yen to make AI more agenticthat is, capable of performing multistep processes on our behalf with some degree of autonomy. A flurry of recent news reflects this trend. On July 9, for example, Perplexity launched Comet, a web browser with a built-in AI agent, available mostly to users of the companys $200/month plan. A week later, OpenAI began rolling out a new ChatGPT agent called . . . Agent. Microsoft is adding a Copilot mode to its Edge browser that it says will soon be able to perform tasks such as making reservations; Opera is previewing Opera Neon, its own browser with built-in agentic AI. Ive been playing with OpenAIs Agent, which showed up in my ChatGPT Plus account earlier this week. The companys blog post on the feature raises expectations by describing it as already a powerful tool for handling complex tasks. So far, however, my experiences with it have not provided any moments of awe and wonder. Instead, Ive been left wondering if the era of offloading all kinds of web work to an LLM is further off than I thought. Tech companies have already trained AI to do some astonishing things, such as achieve gold-medal-level performance in the International Mathematical Olympiad. But Agent often came off like a clueless internet newbie banging its head against a medium conspiring to foil it. In its own odd way, watching Agent at work is fascinating. When you give it a promptas with AI of all types, the more detail you provide, the betterit opens a web browser on a remote OpenAI computer. Then it displays the web pages its accessing right inside your chat and explains every step its taking in absurd detail, down to which buttons it chooses to click. Its like peering into the features brain, and underscores the infinite number of tiny, almost subconscious decisions we make when using the web. [Screenshot: Harry McCracken/ChatGPT] More often than not, though, Agents responses to my requests werent worth the wait. It took 13 minutes to rummage through Google Flights for San Francisco-New York flight options, and the list it gave me was missing the itinerary I probably would have chosen. When I asked it to compile a list of the necessary ingredients to bake authentic German lebkuchen, it combined ones from two different recipes without any apparent logic. I fed it the description for a job opening here at Fast Company and asked it to find candidates; it suggested some, but with out-of-date information on their current employers. After a certain point, I wondered whether the projects I was throwing Agents way were poor tests of its talents. So I tried several tasks ChatGPT suggests when you initiate an Agent session. Many of them, it whiffed. Agent could not log into my Wall Street Journal account to prepare a report on the sites coverage of rare earth materials, or verify my phone number to schedule an Uber pickup. While adding banana cream pie ingredients to an Instacart order, it plugged in a random delivery address and didnt seem to offer any way for me to correct it. A summary of Axioss recent articles on AI worked better, except it didnt include anything from the past two weeks. (Agent was often confused about the current date, informing me at various points that it was July 15 or July 16 when it was actually July 30.) Because Agent discloses what its doing so thoroughly, its possible to hazard some guesses about why the results arent better. First of all, it was frequently bogged down by what it concluded were errors on its part or website malfunctionsIt seems the previous click didnt work as expectedthough it wasnt always clear whether anything had in fact gone wrong. Secondly, the internet as we know it is designed for the convenience of humans, not to facilitate AI agents. Indeed, many sites (including, ahem, FastCompany.com) block automated browsing of the sort Agent performs. In my experience, this blocking was a persistent obstacle to Agent, which kept encountering Are you human? tests. Unfazed, it tried increasingly ambitious work-arounds, such as translating a Fast Company story that had been translated into Spanish back into English. But that turned theoretically simple projects into slogs, almost always with diminishing returns. Lastly, theres the question of privacy and security. Agent is designed to let you type login information for your accounts into its remote browser, though it didnt always work for me. Many folks might be disinclined to even try it, given that it involves handing your passwords over and trusting OpenAI to use them responsibly. In the interest of researching this newsletter, I signed into my Gmail account and asked Agent to compile a few reports on the messages therein. Correctly identifying it as a sensitive situation, Agent insisted I monitor its work and paused it whenever I tabbed awaynegating any time I might have saved by not performing the job myself. Access to the users personal data is essential to Agent realizing even a fraction of its potential, since the better it knows us, the moe sophisticated its help can get. For example, I try to book an aisle seat when flying alone but grab myself a middle seat if my wife is along for the flighta habit a truly clever AI might be able to divine from my travel history without me explicitly stating it. But OpenAI hasnt yet given the feature anything resembling an uncanny ability to understand such needs and desires. For now, Agent often turned out to be a slower way to achieve a goal than existing web tools that are mature and predictable. I was heartened when I asked Agent to find the lowest price on a particular Casio music keyboard: It found it on eBay and added it to my shopping cart. Except that a Google search returned the same eBay listing as its first link. And clicking the Add to cart button oneself does not exactly amount to heavy lifting. The thing is, we already have tools designed to give software, such as an agent, efficient access to other software. Theyre called APIs, and instead of expecting an app to puzzle its way through browsing the web, typing into forms, and clicking forms, they let it transmit requests and retrieve results as streams of raw data. APIs only support processes that the host software has chosen to make available rather than the theoretically open-ended capabilities of an agent. But they do it quickly, easily, and without requiring the user’s attention. Agent does support an existing API-based ChatGPT feature called Connectors, but this, too, was flaky in my experiments. When I issued a Gmail-related request, it didnt point out that there was a Gmail connector but I hadnt installed it. Instead, it had me log into my account and supervise its browsing. Another time, I tried a task involving OneDrive and Agent suggested, fuzzily, that there might be a relevant connector. (There is.) Im not discounting the possibility that Agent, or someone elses agentic web-browsing AI, will get radically better in manifestly obvious ways. Some degree of improvement is inevitable. Yet the tool, in its current state, is another reminder of how far the industrys lofty proclamations have raced ahead of actual progress. OpenAI CEO Sam Altman, Metas Mark Zuckerberg, and others have lately said that their goal is superintelligenceAI thats better than humans at everything. Using a web browser hardly ranks among the worlds most intellectually taxing activities. But until AI masters it, superintelligence will be a talking point, not a reality. Youve been reading Plugged In, Fast Companys weekly tech newsletter from me, global technology editor Harry McCracken. If a friend or colleague forwarded this edition to youor if you’re reading it on FastCompany.comyou can check out previous issues and sign up to get it yourself every Friday morning. I love hearing from you: Ping me at hmccracken@fastcompany.com with your feedback and ideas for future newsletters. I’m also on Bluesky, Mastodon, and Threads, and you can follow Plugged In on Flipboard. More top tech stories from Fast Company How Google is working with Hollywood to bring AI to filmmakingMira Lane, who runs Google’s Envisioning Studio, talks about how artists are embracing tools like Google’s Flow for preproduction, previsualization, and prototyping. Read More Exclusive: Reality Defender expands deepfake detection access to independent developersThe cybersecurity company has launched a public API and a free tier that allows up to 50 detections per month. Read More ‘The overall vibe was total chaos’: Tesla Diner goes viral for long waits and mixed reviewsElon Musk’s retro-futuristic restaurant is blowing up online as TikTokers share their experiences.Read More The Trump administration might overhaul the U.S. patent systemThe Commerce Department is considering charging patent holders between 1% and 5% of their overall patent value. Read More Software nearly ate the world. Now builders and designers are taking it backWhat’s behind the New Industrials movementand why it matters Read More Starbucks was a pioneer of the mobile-first shop. Now its getting rid of themStarbucks is sunsetting its mobile-order and pick-up-only store format as part of a strategy to elevate its café experience. Read More
Category:
E-Commerce
Imagine if Congress had a clear-eyed guide to the technological upheavals shaping our lives. A team of in-house experts who could have flagged the risks of generative AI before ChatGPT went public, raised alarms about deepfakes before they flooded social media, and assessed the vulnerabilities in U.S. infrastructure before ransomware shut down pipelines. For a time, Congress had exactly that, in the form of the Office of Technology Assessment (OTA). But lawmakers shuttered it 30 years ago, and were still feeling its absence today. Created in 1972, the Office of Technology Assessment gave Congress something it almost never has: a reliable way to understand the science and technologies reshaping the world. The offices reports didnt tell lawmakers what to do. Instead, they laid out the risks and the benefits (so cleanly that members on opposite sides of an issue could wave the same report to make their case). The OTA was overseen by a 12member board, split evenly between Democrats and Republicans, with equal representation from the House and the Senate. In just over two decades, it produced over 750 studies, on everything from Alzheimers to automation. It was an impartial repository of interdisciplinary experts who would proactively assist Congress in understanding emerging technology, says University of Washington law professor Ryan Calo, and to do so at a time early enough in its life cycle that it had not become full of special interests that had not grown around it, like barnacles. But not everyone was pleased with OTAs body of work. In 1980, Washington Times reporter Donald Lambro published Fat City: How Washington Wastes Your Taxes, arguing that the agency often focused on issues championed by Senator Ted Kennedy and other liberals. In his view, OTAs studies were duplicative, frequently shoddy, not altogether objective, and often ignored. (Lambros criticisms were, ironically enough, arguably quite partisan: True, OTA sometimes revisited issues already studied by other agencies, but a 1977 Government Accountability Office (GAO) review noted that OTA’s output made “significant contributions in areas of concern to Congress.”) That sentiment carried into the Reagan era. OTAs sharply critical assessments of President Ronald Reagans Strategic Defense Initiative (SDI). Pitched at the height of the Cold War as a revolutionary system of space and groundbased weapons capable of intercepting Soviet missiles in flight, SDI struck supporters as a technological moonshot. OTAs assessment was a splash of cold water: the office warned that the programs staggering cost and ambitious scope offered little assurance it could actually shield the nation from a Soviet attack. Those findings triggered intense political backlash, including from the Heritage Foundation, which in 1984 accused OTA of letting politics override objectivity, claiming that at least one division had prioritized discrediting SDI over providing balanced analysis. The report also argued that flaws in the study and the release of sensitive information were unlikely to be the result of simple mistakes or misunderstanding, concluding: The evidence that some OTA staffers oppose the Administrations Strategic Defense Initiative seems clear and compelling. (Several subsequent independent reviews echoed OTAs assessment of SDI.) The controversy continued when North Carolina Senator Jesse Helms used the SDI dispute to condemn the agency outright. OTA has been obsessed with proving that President Reagans strategic defense initiative is both wrongheaded and dangerous, Helms said in 1988. The political pressure only intensified as the partisan tides shifted. During the 1994 midterm elections, Georgia Representative Newt Gingrich vowed that if Republicans took control of Congress, the Office of Technology Assessment would be on the chopping block. Once his party indeed did sweep into power, Gingrich (now ascended into the role of House Speaker) made good on that promise: In 1995, with a staff of about 140 and an annual budget of roughly $21 million (a rounding error in terms of congressional budgets, Calo says), OTA was quickly defunded, effectively shuttering the office. The move drew swift criticism even from within Gingrichs own party. New York congressman Amo Houghton, for example, lamented, We are cutting off one of the most important arms of Congress when we cut off unbiased knowledge about science and technology. Shuttering OTA solved a partisan problem in 1995, but it left Congress flying blind on science and technology, a gap it has never truly closed. There have been a number of attempts to resurrect OTA, but none have succeeded. House Democrats have floated funding proposals, including a 20192020 effort to allocate $6million to restart the office but these measures died in the Senate. In the meantime, Congress has tried to fill the OTA-sized hole with alternatives like the Government Accountability Offices Science, Technology Assessment, and Analytics (STAA) unit. But while this setup offers some basic technical support, critics argue it lacks OTAs mission-driven focus and deep multidisciplinary expertise, and thus produces far fewer insights than its bureaucratic forebear. They do not have anything like the capacity that the OTA had, says the University of Washington’s Calo. The stakes of that void are becoming increasingly clear. Take, as an example, large language models: An office like OTA could have assessed the risks, outlined guardrails, and prepared Congress before the tools reached the public. Without that kind of early guidance, lawmakers are left reacting after the fact, often leaning on industry lobbyists or outside experts. In the absence of OTA, theres, regrettably, been quite a bit of soft capture by the tech sector, says Jonathan Mayer, a Princeton computer scientist and former Justice Department scienceand technology advisor. And its easy to make the oh, you silly Congress, if only you understood the technology, you’d realize the error of your ways type argument when Congress lacks the technology expertise to respond.” Bruce Schneier, a security technologist and lecturer at Harvard University, argues that the most damning consequence wasnt just the loss of OTA itselfit was what the closure signaled about Congress. It was an early example of ideology trying to shut down facts, he says. And what were left with, he argues, is a tech policy landscape that is shaped largely by lobbyists. Which is not good, he adds, because it comes with an agenda.
Category:
E-Commerce
Even the worlds most bumbling detective could probably figure out why there havent been any comedy blockbusters lately. Amid all the rampaging dinosaurs, broods of superheroes, and live-action Disney characters who technically are not alive, no major comedies whatsoever have graced big screens this summer. (Or last summer.) Thats about to change, though. The Naked Gun reboot careens into theaters this weekend, with Liam Neeson filling in as the films leading man for Leslie Nielsen as shambolic police detective Frank Drebin (this isnt a 007 situationNeesons Drebin is Nielsens son). A spoof franchise thats lain dormant for three decades might not quite sound like a recipe for box office gold (even if it has been a recipe for romance). But Paramount reportedly has a lot of faith in the project, opening it wide against family friendly animated flick The Bad Guys 2. Helmed by Akiva Schaffer, part of the Lonely Island trio with Andy Samberg, The Naked Gun has already won over critics, with reviews hailing it as the funniest movie in years. If it hits hard, it wont just reignite the series, but potentially restore executives faith in theatrically released comedieswhich makes it kind of scary to contemplate what will happen if it flops. When a comedy connects with a packed theater crowd, the energy is electric. Synchronized belly laughs come tumbling out at every turn, seemingly summoned by the rhythm and sheer force of roiling collective joy. Its a visceral, physiological experience, akin to audiences gasping and screaming together at a horror movie. But unlike horror movies, pure comedies rarely make it to the big screen anymoreat least not without modifiers like family-, action-, or indie-. Where have all the comedy blockbusters gone? [Image: Paramount Pictures] When comedy was more common In 2005, Batman Begins kicked off Christopher Nolans Dark Knight trilogy. It was a massive hit, propelling Christian Bale to stardom and helping usher in the modern era of superhero movies. What now seems lost to history about the films success, however, is that it was slightly exceeded by the Vince Vaughn goof-fest Wedding Crashers, which earned $209 million to Batman Beginss $206 million at the domestic box officeon roughly a quarter of the budget. Wedding Crashers was an unusually big comedy 20 years ago, but not that unusual. It was one of 28 from the genre to reach the top 100 highest-grossers that yeara list that, in domestic box office returns, includes Will Smiths Hitch ($179 million), the Ben Stiller versus Robert De Niro sequel Meet the Fockers ($146 million), and The 40-Year-Old Virgin ($109 million), which kicked off Judd Apatows reign as something of a raunch-comedy auteur. The following year had even more comedies in the top 100, including Borat ($128 million) and Will Ferrells Talladega Nights ($148 million). Typically cheaper to make than special-effects-driven blockbusters, comedies remained a wildly profitable part of the box office throughout the early 2010s. As established stars like Vaughn and Adam Sandler began shedding some of their box office potency, new comedic heroes like Zach Galifianakis (The Hangover, $277 million) and Melissa McCarthy (The Heat, $159 million) emerged, right on schedule, to pick up the slack. By the back half of the decade, though, there seemed to be nothing but slack. [Photo: Frank Masi/ 2025 Paramount Pictures] The case of the vanishing chuckles Sandler saw the writing on the wall fairly early. In 2014the same year that Blended, his reunion with Wedding Singer costar Drew Barrymore, grossed a relatively paltry $46 millionSandler signed his first deal with Netflix to make comedies that live exclusively on the streaming platform. It was an ominous, accurate indicator that original streaming content was coming to eat the box offices lunchand comedies appeared to be first on the menu. The revolution rolled in gradually. In 2015, Apatow had his last big hit to date with Trainwreck, which made $110 million and anointed Amy Schumer as the next major comedy movie star. Tiffany Haddish snatched that title two years later, as the breakout of Girls Trip ($115 million), but neither star managed to sustain their momentum, with returns quickly diminishing. Meanwhile, the ceiling for big-screen comedies collapsed, as well-regarded hits like 2018s Game Night and Blockers topped out between $60 million and $70 million. (Crazy Rich Asians made a crazy rich $174 million that summer, but its more of a romantic comedy/drama than a pure comedy.) Buoyed by the dominance of superhero fare like Avengers: Endgame ($858 million domestic), 2019 became the biggest box office year on record, but comedy had comparatively meager returnscementing the conventional wisdom that the genre was no longer profitable. It coincided with the arrival of fresh streamers like Apple TV+ and Peacock, hungry for exclusive content, and COVID-19 keeping millions of moviegoers out of theaters for years. As far as theatrical releases go, comedy is now mostly relegated to an element that filmmakers inject into other genres. Thor: Love and Thunder ($343 million), is now a comedy. M3GAN ($95 million), with its made-to-be-memed robot dance, is a comedy. Barbie ($636 million) with its money-printing IP and the black swan event-iness of opening against Oppenheimer, is a comedy. As for pure comedies from legendary stars of the genre like Eddie Murphy, though, one need only scroll through Netflix and Prime Video, where movies like You People and Coming 2 America exclusively reside. Big-screen comedies, it would seem, are cooked. But maybe they dont have to be. Whats riding on The Naked Gun hitting The problem facing theatrical comedies now is one of expectations. As the potential for staggering worldwide grosses grew over the past two decades, studios became interested in top global performers and little elseand comedy tends not to translate well. Everything in theaters now seems to either be a $500 million juggernaut or a $20 million A24 movie made for pocket change. Whatever happened to midbudget movies and mid-level expectations? One of the last hyped comedies to hit theaters, the 2023 Jennifer Lawrence vehicle No Hard Feelings, was considered a flop because it only made $50 million domestically on a $45 million budget. This years One of Them Days, however, starring Keke Palmer and SZA, took in the same total on a third of the budget, and is considered a modest hit. (A sequel is already in the works.) If audiences will turn out enough to make that comedy profitable, it stands to reason theres a lot of untapped potential left in the genre. With a reported $42 million budget, The Naked Gun will have to shoot a little higher to be considered a hit. If it manages to clear that magic $100 mark, though, something a pure comedy hasnt done in ages, it could remind the decision-makers that filmgoers love to laugh. It could lay the groundwork for a big marketing push for Aziz Ansaris filmmaking debut, Good Fortune, starring Seth Rogen, which is slated for this fall. (I dont buy into this theory that theatrical comedies are not a thing, Ansari has said about his hopes for the film.) It might not take more than one or two major hit comedies before the genre is deemed theatrically viable again. If The Naked Gun flops, however, comedy stands to be further relegated to the world of streaming, where the stakes are low and so, too often, is the level of effort. Theres nothing funny about that.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||