This week, we published Power Hungry, a package all about AI and energy. At the center of this package is the most comprehensive look yet at AI’s growing power demand, if I do say so myself.  This data-heavy story is the result of over six months of reporting by me and my colleague James O’Donnell (and the work of many others on our team). Over that time, with the help of leading researchers, we quantified the energy and emissions impacts of individual queries to AI models and tallied what it all adds up to, both right now and for the years ahead.  There’s a lot of data to dig through, and I hope you’ll take the time to explore the whole story. But in the meantime, here are three of my biggest takeaways from working on this project.  1. The energy demands of AI are anything but constant.  If you’ve heard estimates of AI’s toll, it’s probably a single number associated with a query, likely to OpenAI’s ChatGPT. One popular estimate is that writing an email with ChatGPT uses 500 milliliters (or roughly a bottle) of water. But as we started reporting, I was surprised to learn just how much the details of a query can affect its energy demand. No two queries are the same—for several reasons, including their complexity and the particulars of the model being queried. One key caveat here is that we don’t know much about “closed source” models—for these, companies hold back the details of how they work. (OpenAI’s ChatGPT and Google’s Gemini are examples.) Instead, we worked with researchers who measured the energy it takes to run open-source AI models, for which the source code is publicly available.  But using open-source models, it’s possible to directly measure the energy used to respond to a query rather than just guess. We worked with researchers who generated text, images, and video and measured the energy required for the chips the models are based on to perform the task.   Even just within the text responses, there was a pretty large range of energy needs. A complicated travel itinerary consumed nearly 10 times as much energy as a simple request for a few jokes, for example. An even bigger difference comes from the size of the model used. Larger models with more parameters used up to 70 times more energy than smaller ones for the same prompts.  As you might imagine, there’s also a big difference between text, images, or video. Videos generally took hundreds of times more energy to generate than text responses.  2. What’s powering the grid will greatly affect the climate toll of AI’s energy use.  As the resident climate reporter on this project, I was excited to take the expected energy toll and translate it into an expected emissions burden.  Powering a data center with a nuclear reactor or a whole bunch of solar panels and batteries will not affect our planet the same way as burning mountains of coal. To quantify this idea, we used a figure called carbon intensity, a measure of how dirty a unit of electricity is on a given grid.  We found that the same exact query, with the same exact energy demand, will have a very different climate impact depending on what the data center is powered by, and that depends on the location and the time of day. For example, querying a data center in West Virginia could cause nearly twice the emissions of querying one in California, according to calculations based on average data from 2024. This point shows why it matters where tech giants are building data centers, what the grid looks like in their chosen locations, and how that might change with more demand from the new infrastructure.  3. There is still so much that we don't know when it comes to AI and energy.  Our reporting resulted in estimates that are some of the most specific and comprehensive out there. But ultimately, we still have no idea what many of the biggest, most influential models are adding up to in terms of energy and emissions. None of the companies we reached out to were willing to provide numbers during our reporting. Not one. Adding up our estimates can only go so far, in part because AI is increasingly everywhere. While today you might generally have to go to a dedicated site and type in questions, in the future AI could be stitched into the fabric of our interactions with technology. (See my colleague Will Douglas Heaven’s new story on Google’s I/O showcase: “By putting AI into everything, Google wants to make it invisible.”) AI could be one of the major forces that shape our society, our work, and our power grid. Knowing more about its consequences could be crucial to planning our future.  To dig into our reporting, give the main story a read. And if you’re looking for more details on how we came up with our numbers, you can check out this behind-the-scenes piece. There are also some great related stories in this package, including one from James Temple on the data center boom in the Nevada desert, one from David Rotman about how AI’s rise could entrench natural gas, and one from Will Douglas Heaven on a few technical innovations that could help make AI more efficient. Oh, and I also have a piece on why nuclear isn’t the easy answer some think it is.  Find them, and the rest of the stories in the package, here.  This article is from The Spark, MIT Technology Review’s weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here.

Honeybees may use an electric sense to locate nectar-rich flowersproxyminder/Getty Images Electric signals from power lines discourage bees from landing on nearby flowers, and there is growing concern that this so-called “electric pollution” could be causing widespread disruption to insect behaviour. Alongside bumblebees, honeybees have been shown to detect and respond to airborne electric fields – which are often caused by static electricity in the natural world – detected through hairs or antennae. Research has shown that honeybees use electro-reception as a form of communication, and both honeybees and bumblebees are thought to use it to locate nectar-rich flowers. Read more…

This week, we published Power Hungry, a package all about AI and energy. At the center of this package is the most comprehensive look yet at AI’s growing power demand, if I do say so myself.  This data-heavy story is the result of over six months of reporting by me and my colleague James O’Donnell (and the work of many others on our team). Over that time, with the help of leading researchers, we quantified the energy and emissions impacts of individual queries to AI models and tallied what it all adds up to, both right now and for the years ahead.  There’s a lot of data to dig through, and I hope you’ll take the time to explore the whole story. But in the meantime, here are three of my biggest takeaways from working on this project.  1. The energy demands of AI are anything but constant.  If you’ve heard estimates of AI’s toll, it’s probably a single number associated with a query, likely to OpenAI’s ChatGPT. One popular estimate is that writing an email with ChatGPT uses 500 milliliters (or roughly a bottle) of water. But as we started reporting, I was surprised to learn just how much the details of a query can affect its energy demand. No two queries are the same—for several reasons, including their complexity and the particulars of the model being queried. One key caveat here is that we don’t know much about “closed source” models—for these, companies hold back the details of how they work. (OpenAI’s ChatGPT and Google’s Gemini are examples.) Instead, we worked with researchers who measured the energy it takes to run open-source AI models, for which the source code is publicly available.  But using open-source models, it’s possible to directly measure the energy used to respond to a query rather than just guess. We worked with researchers who generated text, images, and video and measured the energy required for the chips the models are based on to perform the task.   Even just within the text responses, there was a pretty large range of energy needs. A complicated travel itinerary consumed nearly 10 times as much energy as a simple request for a few jokes, for example. An even bigger difference comes from the size of the model used. Larger models with more parameters used up to 70 times more energy than smaller ones for the same prompts.  As you might imagine, there’s also a big difference between text, images, or video. Videos generally took hundreds of times more energy to generate than text responses.  2. What’s powering the grid will greatly affect the climate toll of AI’s energy use.  As the resident climate reporter on this project, I was excited to take the expected energy toll and translate it into an expected emissions burden.  Powering a data center with a nuclear reactor or a whole bunch of solar panels and batteries will not affect our planet the same way as burning mountains of coal. To quantify this idea, we used a figure called carbon intensity, a measure of how dirty a unit of electricity is on a given grid.  We found that the same exact query, with the same exact energy demand, will have a very different climate impact depending on what the data center is powered by, and that depends on the location and the time of day. For example, querying a data center in West Virginia could cause nearly twice the emissions of querying one in California, according to calculations based on average data from 2024. This point shows why it matters where tech giants are building data centers, what the grid looks like in their chosen locations, and how that might change with more demand from the new infrastructure.  3. There is still so much that we don’t know when it comes to AI and energy.  Our reporting resulted in estimates that are some of the most specific and comprehensive out there. But ultimately, we still have no idea what many of the biggest, most influential models are adding up to in terms of energy and emissions. None of the companies we reached out to were willing to provide numbers during our reporting. Not one. Adding up our estimates can only go so far, in part because AI is increasingly everywhere. While today you might generally have to go to a dedicated site and type in questions, in the future AI could be stitched into the fabric of our interactions with technology. (See my colleague Will Douglas Heaven’s new story on Google’s I/O showcase: “By putting AI into everything, Google wants to make it invisible.”) AI could be one of the major forces that shape our society, our work, and our power grid. Knowing more about its consequences could be crucial to planning our future.  To dig into our reporting, give the main story a read. And if you’re looking for more details on how we came up with our numbers, you can check out this behind-the-scenes piece. There are also some great related stories in this package, including one from James Temple on the data center boom in the Nevada desert, one from David Rotman about how AI’s rise could entrench natural gas, and one from Will Douglas Heaven on a few technical innovations that could help make AI more efficient. Oh, and I also have a piece on why nuclear isn’t the easy answer some think it is.  Find them, and the rest of the stories in the package, here.  This article is from The Spark, MIT Technology Review’s weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here.

This week, we published Power Hungry, a package all about AI and energy. At the center of this package is the most comprehensive look yet at AI’s growing power demand, if I do say so myself.  This data-heavy story is the result of over six months of reporting by me and my colleague James O’Donnell (and the work of many others on our team). Over that time, with the help of leading researchers, we quantified the energy and emissions impacts of individual queries to AI models and tallied what it all adds up to, both right now and for the years ahead.  There’s a lot of data to dig through, and I hope you’ll take the time to explore the whole story. But in the meantime, here are three of my biggest takeaways from working on this project.  1. The energy demands of AI are anything but constant.  If you’ve heard estimates of AI’s toll, it’s probably a single number associated with a query, likely to OpenAI’s ChatGPT. One popular estimate is that writing an email with ChatGPT uses 500 milliliters (or roughly a bottle) of water. But as we started reporting, I was surprised to learn just how much the details of a query can affect its energy demand. No two queries are the same—for several reasons, including their complexity and the particulars of the model being queried. One key caveat here is that we don’t know much about “closed source” models—for these, companies hold back the details of how they work. (OpenAI’s ChatGPT and Google’s Gemini are examples.) Instead, we worked with researchers who measured the energy it takes to run open-source AI models, for which the source code is publicly available.  But using open-source models, it’s possible to directly measure the energy used to respond to a query rather than just guess. We worked with researchers who generated text, images, and video and measured the energy required for the chips the models are based on to perform the task.   Even just within the text responses, there was a pretty large range of energy needs. A complicated travel itinerary consumed nearly 10 times as much energy as a simple request for a few jokes, for example. An even bigger difference comes from the size of the model used. Larger models with more parameters used up to 70 times more energy than smaller ones for the same prompts.  As you might imagine, there’s also a big difference between text, images, or video. Videos generally took hundreds of times more energy to generate than text responses.  2. What’s powering the grid will greatly affect the climate toll of AI’s energy use.  As the resident climate reporter on this project, I was excited to take the expected energy toll and translate it into an expected emissions burden.  Powering a data center with a nuclear reactor or a whole bunch of solar panels and batteries will not affect our planet the same way as burning mountains of coal. To quantify this idea, we used a figure called carbon intensity, a measure of how dirty a unit of electricity is on a given grid.  We found that the same exact query, with the same exact energy demand, will have a very different climate impact depending on what the data center is powered by, and that depends on the location and the time of day. For example, querying a data center in West Virginia could cause nearly twice the emissions of querying one in California, according to calculations based on average data from 2024. This point shows why it matters where tech giants are building data centers, what the grid looks like in their chosen locations, and how that might change with more demand from the new infrastructure.  3. There is still so much that we don't know when it comes to AI and energy.  Our reporting resulted in estimates that are some of the most specific and comprehensive out there. But ultimately, we still have no idea what many of the biggest, most influential models are adding up to in terms of energy and emissions. None of the companies we reached out to were willing to provide numbers during our reporting. Not one. Adding up our estimates can only go so far, in part because AI is increasingly everywhere. While today you might generally have to go to a dedicated site and type in questions, in the future AI could be stitched into the fabric of our interactions with technology. (See my colleague Will Douglas Heaven’s new story on Google’s I/O showcase: “By putting AI into everything, Google wants to make it invisible.”) AI could be one of the major forces that shape our society, our work, and our power grid. Knowing more about its consequences could be crucial to planning our future.  To dig into our reporting, give the main story a read. And if you’re looking for more details on how we came up with our numbers, you can check out this behind-the-scenes piece. There are also some great related stories in this package, including one from James Temple on the data center boom in the Nevada desert, one from David Rotman about how AI’s rise could entrench natural gas, and one from Will Douglas Heaven on a few technical innovations that could help make AI more efficient. Oh, and I also have a piece on why nuclear isn’t the easy answer some think it is.  Find them, and the rest of the stories in the package, here.  This article is from The Spark, MIT Technology Review’s weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here.

Content warning: this story includes discussion of self-harm and suicide. If you are in crisis, please call, text or chat with the Suicide and Crisis Lifeline at 988, or contact the Crisis Text Line by texting TALK to 741741.A judge in Florida just rejected a motion to dismiss a lawsuit alleging that the chatbot startup Character.AI — and its closely tied benefactor, Google — caused the death by suicide of a 14-year-old user, clearing the way for the first-of-its-kind lawsuit to move forward in court.The lawsuit, filed in October, claims that recklessly released Character.AI chatbots sexually and emotionally abused a teenage user, Sewell Setzer III, resulting in obsessive use of the platform, mental and emotional suffering, and ultimately his suicide in February 2024.In January, the defendants in the case — Character.AI, Google, and Character.AI cofounders Noam Shazeer and Daniel de Freitas — filed a motion to dismiss the case mainly on First Amendment grounds, arguing that AI-generated chatbot outputs qualify as speech, and that "allegedly harmful speech, including speech allegedly resulting in suicide," is protected under the First Amendment.But this argument didn't quite cut it, the judge ruled, at least not in this early stage. In her opinion, presiding US district judge Anne Conway said the companies failed to sufficiently show that AI-generated outputs produced by large language models (LLMs) are more than simply words — as opposed to speech, which hinges on intent.The defendants "fail to articulate," Conway wrote in her ruling, "why words strung together by an LLM are speech."The motion to dismiss did find some success, with Conway dismissing specific claims regarding the alleged "intentional infliction of emotional distress," or IIED. (It's difficult to prove IIED when the person who allegedly suffered it, in this case Setzer, is no longer alive.)Still, the ruling is a blow to the high-powered Silicon Valley defendants who had sought to have the suit tossed out entirely.Significantly, Conway's opinion allows Megan Garcia, Setzer's mother and the plaintiff in the case, to sue Character.AI, Google, Shazeer, and de Freitas on product liability grounds. Garcia and her lawyers argue that Character.AI is a product, and that it was rolled out recklessly to the public, teens included, despite known and possibly destructive risks.In the eyes of the law, tech companies generally prefer to see their creations as services, like electricity or the internet, rather than products, like cars or nonstick frying pans. Services can't be held accountable for product liability claims, including claims of negligence, but products can.In a statement, Tech Justice Law Project director and founder Meetali Jain, who's co-counsel for Garcia alongside Social Media Victims Law Center founder Matt Bergman, celebrated the ruling as a win — not just for this particular case, but for tech policy advocates writ large."With today's ruling, a federal judge recognizes a grieving mother's right to access the courts to hold powerful tech companies — and their developers — accountable for marketing a defective product that led to her child's death," said Jain."This historic ruling not only allows Megan Garcia to seek the justice her family deserves," Jain added, "but also sets a new precedent for legal accountability across the AI and tech ecosystem."Character.AI was founded by Shazeer and de Freitas in 2021; the duo had worked together on AI projects at Google, and left together to launch their own chatbot startup. Google provided Character.AI with its essential Cloud infrastructure, and in 2024 raised eyebrows when it paid Character.AI $2.7 billion to license the chatbot firm's data — and bring its cofounders, as well as 30 other Character.AI staffers, into Google's fold. Shazeer, in particular, now holds a hugely influential position at Google DeepMind, where he serves as a VP and co-lead for Google's Gemini LLM.Google did not respond to a request for comment at the time of publishing, but a spokesperson for the search giant told Reuters that Google and Character.AI are "entirely separate" and that Google "did not create, design, or manage" the Character.AI app "or any component part of it."In a statement, a spokesperson for Character.AI emphasized recent safety updates issued following the news of Garcia's lawsuit, and said it "looked forward" to its continued defense:It's long been true that the law takes time to adapt to new technology, and AI is no different. In today's order, the court made clear that it was not ready to rule on all of Character.AI 's arguments at this stage and we look forward to continuing to defend the merits of the case.We care deeply about the safety of our users and our goal is to provide a space that is engaging and safe. We have launched a number of safety features that aim to achieve that balance, including a separate version of our Large Language Model model for under-18 users, parental insights, filtered Characters, time spent notification, updated prominent disclaimers and more.Additionally, we have a number of technical protections aimed at detecting and preventing conversations about self-harm on the platform; in certain cases, that includes surfacing a specific pop-up directing users to the National Suicide and Crisis Lifeline.Any safety-focused changes, though, were made months after Setzer's death and after the eventual filing of the lawsuit, and can't apply to the court's ultimate decision in the case.Meanwhile, journalists and researchers continue to find holes in the chatbot site's upxdated safety protocols. Weeks after news of the lawsuit was announced, for example, we continued to find chatbots expressly dedicated to self-harm, grooming and pedophilia, eating disorders, and mass violence. And a team of researchers, including psychologists at Stanford, recently found that using a Character.AI voice feature called "Character Calls" effectively nukes any semblance of guardrails — and determined that no kid under 18 should be using AI companions, including Character.AI.Share This Article

New analysis by MIT Technology Review reveals AI's rapidly growing energy demands, with data centers expected to triple their share of US electricity consumption from 4.4% to 12% by 2028. According to Lawrence Berkeley National Laboratory projections, AI alone could soon consume electricity equivalent to 22% of all US households annually, driven primarily by inference operations that represent 80-90% of AI's computing power. The carbon intensity of electricity used by data centers is 48% higher than the US average, researchers found, as facilities increasingly turn to dirtier energy sources like natural gas to meet immediate needs. Tech giants are racing to secure unprecedented energy resources: OpenAI and President Trump announced a $500 billion Stargate initiative, Apple plans to spend $500 billion on manufacturing and data centers, and Google expects to invest $75 billion in AI infrastructure in 2025 alone. Despite their massive energy ambitions, leading AI companies remain largely silent about their per-query energy consumption, leaving researchers struggling to assemble what one expert called "a total black box." Read more of this story at Slashdot.

News Physics Scientists used a levitating magnet to hunt for dark matter The quantum-based technique could reveal the presence of hypothetical ultralight particles A magnet levitates over a superconductor. Scientists used this phenomenon in a search for ultralight dark matter. Forance/Alamy Stock Photo By Emily Conover 1 hour ago In a first-of-its-kind test, scientists used a levitated magnet to search for dark matter, the unidentified substance believed to be present throughout the cosmos. If dark matter is made up of ultralight particles, it could behave like a wave that would subtly jostle the magnet. Although no signs of such jostling appeared, a few tweaks could improve the experiment’s sensitivity to dark matter’s potential influence, the researchers report in a paper to be published in Physical Review Letters. Scientists used this apparatus to make measurements of a levitated magnet that provided a new method to hunt for dark matterDennis Uitenbroekz Scientists believe dark matter must exist to account for astronomical observations that suggest an unidentified source of mass in the cosmos. Dark matter has evaded a wide variety of detection attempts, so astroparticle physicist Christopher Tunnell of Rice University in Houston was looking for new ways to search for it. Tunnell and colleagues realized that an existing experiment, originally designed to make sensitive gravitational measurements, could spot ultralight dark matter. In the experiment, a magnet with a mass of less than a milligram was suspended within a container made of a superconductor — a material that conducts electricity without resistance. The magnet’s motion was monitored with a quantum device designed to measures changes in magnetic fields, known as a SQUID. If ultralight dark matter exists and interacts with normal matter via a new type of force, the dark matter wave would have subtly jiggled the magnet. Tunnell and colleagues reanalyzed the data from that experiment to search for dark matter’s influence, but found no evidence of it. Sign up for our newsletter We summarize the week's scientific breakthroughs every Thursday.

This is today's edition of The Download, our weekday newsletter that provides a daily dose of what's going on in the world of technology. By putting AI into everything, Google wants to make it invisible If you want to know where AI is headed, this year’s Google I/O has you covered. The company’s annual showcase of next-gen products, which kicked off yesterday, has all of the pomp and pizzazz, the sizzle reels and celebrity walk-ons, that you’d expect from a multimillion dollar marketing event. But it also shows us just how fast this still-experimental technology is being subsumed into a line-up designed to sell phones and subscription tiers. Never before have I seen this thing we call artificial intelligence appear so normal. Read the full story.—Will Douglas Heaven AI could keep us dependent on natural gas for decades to come Last December, Meta announced plans to build a massive $10 billion data center for training its artificial intelligence models in rural northeast Louisiana. Stretching for more than a mile, it will be Meta’s largest in the world, and it will have an enormous appetite for electricity. To power the data center, a Meta contractor called Entergy will build three large natural-gas power plants with a total capacity of 2.3 gigawatts. It’ll also upgrade the grid to accommodate the huge jump in anticipated demand.The choice of natural gas as the go-to solution to meet the growing demand for power from AI is not unique to Louisiana. The fossil fuel is already the country’s chief source of electricity generation, and large natural-gas plants are being built around the country to feed electricity to new and planned AI data centers. That’s all but wiping out any prospect that the US will wean itself off natural gas anytime soon. Read the full story.—David Rotman This story is part of Power Hungry: AI and our energy future—our new series shining a light on AI’s energy usage. Check out the rest of the package here. Take a new look at AI’s energy use Big Tech's appetite for energy is growing rapidly as adoption of AI accelerates. But just how much energy does a single AI query use? And what does it mean for the climate?  Join editor in chief Mat Honan, senior climate reporter Casey Crownhart, and AI reporter James O'Donnell at 1.30pm ET today for a subscriber-only Roundtables conversation digging into our new package of stories about AI's energy demands now and in the future. Register here.The must-reads I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology. 1 Democrats are on the hunt for a digital thought leader They’re (finally) realizing how far they’re lagging behind their opponents’ online efforts these days. (NYT $) + AI’s impact on elections is being overblown. (MIT Technology Review)2 At least two newspapers printed an AI-generated summer reading list 📰 The only problem is, some of the books don’t actually exist. (404 Media)+ It’s a useful reminder to never take anything chatbots produce as fact. (Axios)+ Even regional newspapers aren’t safe from AI slop. (The Atlantic $)+ Why AI hallucinates, and why we can’t stop it. (MIT Technology Review)3 The Earth may already be too hot to maintain polar ice sheetsEven if it stays at current temperature levels. (WP $)+ Why climate researchers are taking the temperature of mountain snow. (MIT Technology Review)4 How New York City’s child abuse algorithm flags families for investigation Critics believe it’s open to racial bias. (The Markup) 5 Here’s what it’s like to interview for a job at DOGE   The hiring process is remarkably fast, for a government entity. (Wired $)+ The department reportedly tried to enter the US government’s publishing operation. (Politico)+ DOGE’s tech takeover threatens the safety and stability of our critical data. (MIT Technology Review)6 Fortnite has finally returned to Apple’s App StoreAfter five years and a lengthy legal battle. (NYT $) + The recent ruling has major implications for the iOS economy. (Reuters)7 Most chatbots can be tricked into dispensing dangerous informationFrom hacking advice, to describing how to make drugs. (The Guardian) + Anthropic has a new way to protect large language models against jailbreaks. (MIT Technology Review)8 Young Indonesians are being trafficked to scam farmsFraudulent job ads on Telegram and Facebook lure them into a life of crime. (Rest of World) + Inside a romance scam compound—and how people get tricked into being there. (MIT Technology Review) 9  Inside the building in China where stolen western iPhones are stripped and sold You’ll find a buyer for every single component inside the Feiyang Times. (FT $)10 Amazon has started randomly refunding customers for old purchasesSome orders were placed as far back as 2018. (Bloomberg $) Quote of the day “Anybody who’s a computer scientist should not be retired right now. They should be working on AI.” —Google cofounder Sergey Brin says people with the right technical skills should copy him and quit being retired, TechCrunch reports. One more thing This fuel plant will use agricultural waste to combat climate change A startup called Mote plans to build a new type of fuel-producing plant in California’s fertile Central Valley that would, if it works as hoped, continually capture and bury carbon dioxide.It’s among a growing number of efforts to commercialize a concept first proposed two decades ago as a means of combating climate change, known as bioenergy with carbon capture and sequestration, or BECCS.It’s an ambitious plan. However, there are serious challenges to doing BECCS affordably and in ways that reliably suck down significant levels of carbon dioxide. Read the full story.—James Temple We can still have nice things A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet 'em at me.) + These creepy little Labubu toys are everywhere. But why?+ Happy 25th birthday to one of London’s finest institutions, the Tate Modern gallery.+ Why the Mission Impossible film franchise just won’t die.+ Hummingbirds can fly backwards!? Wow.

This is today's edition of The Download, our weekday newsletter that provides a daily dose of what's going on in the world of technology. By putting AI into everything, Google wants to make it invisible If you want to know where AI is headed, this year’s Google I/O has you covered. The company’s annual showcase of next-gen products, which kicked off yesterday, has all of the pomp and pizzazz, the sizzle reels and celebrity walk-ons, that you’d expect from a multimillion dollar marketing event. But it also shows us just how fast this still-experimental technology is being subsumed into a line-up designed to sell phones and subscription tiers. Never before have I seen this thing we call artificial intelligence appear so normal. Read the full story.—Will Douglas Heaven AI could keep us dependent on natural gas for decades to come Last December, Meta announced plans to build a massive $10 billion data center for training its artificial intelligence models in rural northeast Louisiana. Stretching for more than a mile, it will be Meta’s largest in the world, and it will have an enormous appetite for electricity. To power the data center, a Meta contractor called Entergy will build three large natural-gas power plants with a total capacity of 2.3 gigawatts. It’ll also upgrade the grid to accommodate the huge jump in anticipated demand.The choice of natural gas as the go-to solution to meet the growing demand for power from AI is not unique to Louisiana. The fossil fuel is already the country’s chief source of electricity generation, and large natural-gas plants are being built around the country to feed electricity to new and planned AI data centers. That’s all but wiping out any prospect that the US will wean itself off natural gas anytime soon. Read the full story.—David Rotman This story is part of Power Hungry: AI and our energy future—our new series shining a light on AI’s energy usage. Check out the rest of the package here. Take a new look at AI’s energy use Big Tech's appetite for energy is growing rapidly as adoption of AI accelerates. But just how much energy does a single AI query use? And what does it mean for the climate?  Join editor in chief Mat Honan, senior climate reporter Casey Crownhart, and AI reporter James O'Donnell at 1.30pm ET today for a subscriber-only Roundtables conversation digging into our new package of stories about AI's energy demands now and in the future. Register here.The must-reads I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology. 1 Democrats are on the hunt for a digital thought leader They’re (finally) realizing how far they’re lagging behind their opponents’ online efforts these days. (NYT $) + AI’s impact on elections is being overblown. (MIT Technology Review)2 At least two newspapers printed an AI-generated summer reading list 📰 The only problem is, some of the books don’t actually exist. (404 Media)+ It’s a useful reminder to never take anything chatbots produce as fact. (Axios)+ Even regional newspapers aren’t safe from AI slop. (The Atlantic $)+ Why AI hallucinates, and why we can’t stop it. (MIT Technology Review)3 The Earth may already be too hot to maintain polar ice sheetsEven if it stays at current temperature levels. (WP $)+ Why climate researchers are taking the temperature of mountain snow. (MIT Technology Review)4 How New York City’s child abuse algorithm flags families for investigation Critics believe it’s open to racial bias. (The Markup) 5 Here’s what it’s like to interview for a job at DOGE   The hiring process is remarkably fast, for a government entity. (Wired $)+ The department reportedly tried to enter the US government’s publishing operation. (Politico)+ DOGE’s tech takeover threatens the safety and stability of our critical data. (MIT Technology Review)6 Fortnite has finally returned to Apple’s App StoreAfter five years and a lengthy legal battle. (NYT $) + The recent ruling has major implications for the iOS economy. (Reuters)7 Most chatbots can be tricked into dispensing dangerous informationFrom hacking advice, to describing how to make drugs. (The Guardian) + Anthropic has a new way to protect large language models against jailbreaks. (MIT Technology Review)8 Young Indonesians are being trafficked to scam farmsFraudulent job ads on Telegram and Facebook lure them into a life of crime. (Rest of World) + Inside a romance scam compound—and how people get tricked into being there. (MIT Technology Review) 9  Inside the building in China where stolen western iPhones are stripped and sold You’ll find a buyer for every single component inside the Feiyang Times. (FT $)10 Amazon has started randomly refunding customers for old purchasesSome orders were placed as far back as 2018. (Bloomberg $) Quote of the day “Anybody who’s a computer scientist should not be retired right now. They should be working on AI.” —Google cofounder Sergey Brin says people with the right technical skills should copy him and quit being retired, TechCrunch reports. One more thing This fuel plant will use agricultural waste to combat climate change A startup called Mote plans to build a new type of fuel-producing plant in California’s fertile Central Valley that would, if it works as hoped, continually capture and bury carbon dioxide.It’s among a growing number of efforts to commercialize a concept first proposed two decades ago as a means of combating climate change, known as bioenergy with carbon capture and sequestration, or BECCS.It’s an ambitious plan. However, there are serious challenges to doing BECCS affordably and in ways that reliably suck down significant levels of carbon dioxide. Read the full story.—James Temple We can still have nice things A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet 'em at me.) + These creepy little Labubu toys are everywhere. But why?+ Happy 25th birthday to one of London’s finest institutions, the Tate Modern gallery.+ Why the Mission Impossible film franchise just won’t die.+ Hummingbirds can fly backwards!? Wow.

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. By putting AI into everything, Google wants to make it invisible If you want to know where AI is headed, this year’s Google I/O has you covered. The company’s annual showcase of next-gen products, which kicked off yesterday, has all of the pomp and pizzazz, the sizzle reels and celebrity walk-ons, that you’d expect from a multimillion dollar marketing event.But it also shows us just how fast this still-experimental technology is being subsumed into a line-up designed to sell phones and subscription tiers. Never before have I seen this thing we call artificial intelligence appear so normal. Read the full story. —Will Douglas Heaven AI could keep us dependent on natural gas for decades to come Last December, Meta announced plans to build a massive $10 billion data center for training its artificial intelligence models in rural northeast Louisiana. Stretching for more than a mile, it will be Meta’s largest in the world, and it will have an enormous appetite for electricity.To power the data center, a Meta contractor called Entergy will build three large natural-gas power plants with a total capacity of 2.3 gigawatts. It’ll also upgrade the grid to accommodate the huge jump in anticipated demand.The choice of natural gas as the go-to solution to meet the growing demand for power from AI is not unique to Louisiana. The fossil fuel is already the country’s chief source of electricity generation, and large natural-gas plants are being built around the country to feed electricity to new and planned AI data centers. That’s all but wiping out any prospect that the US will wean itself off natural gas anytime soon. Read the full story. —David Rotman This story is part of Power Hungry: AI and our energy future—our new series shining a light on AI’s energy usage. Check out the rest of the package here. Take a new look at AI’s energy use Big Tech’s appetite for energy is growing rapidly as adoption of AI accelerates. But just how much energy does a single AI query use? And what does it mean for the climate? Join editor in chief Mat Honan, senior climate reporter Casey Crownhart, and AI reporter James O’Donnell at 1.30pm ET today for a subscriber-only Roundtables conversation digging into our new package of stories about AI’s energy demands now and in the future. Register here. The must-reads I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology. 1 Democrats are on the hunt for a digital thought leaderThey’re (finally) realizing how far they’re lagging behind their opponents’ online efforts these days. (NYT $) + AI’s impact on elections is being overblown. (MIT Technology Review) 2 At least two newspapers printed an AI-generated summer reading list The only problem is, some of the books don’t actually exist. (404 Media)+ It’s a useful reminder to never take anything chatbots produce as fact. (Axios)+ Even regional newspapers aren’t safe from AI slop. (The Atlantic $)+ Why AI hallucinates, and why we can’t stop it. (MIT Technology Review)3 The Earth may already be too hot to maintain polar ice sheetsEven if it stays at current temperature levels. (WP $)+ Why climate researchers are taking the temperature of mountain snow. (MIT Technology Review) 4 How New York City’s child abuse algorithm flags families for investigationCritics believe it’s open to racial bias. (The Markup) 5 Here’s what it’s like to interview for a job at DOGE  The hiring process is remarkably fast, for a government entity. (Wired $)+ The department reportedly tried to enter the US government’s publishing operation. (Politico)+ DOGE’s tech takeover threatens the safety and stability of our critical data. (MIT Technology Review) 6 Fortnite has finally returned to Apple’s App StoreAfter five years and a lengthy legal battle. (NYT $)+ The recent ruling has major implications for the iOS economy. (Reuters) 7 Most chatbots can be tricked into dispensing dangerous informationFrom hacking advice, to describing how to make drugs. (The Guardian)+ Anthropic has a new way to protect large language models against jailbreaks. (MIT Technology Review) 8 Young Indonesians are being trafficked to scam farmsFraudulent job ads on Telegram and Facebook lure them into a life of crime. (Rest of World)+ Inside a romance scam compound—and how people get tricked into being there. (MIT Technology Review) 9  Inside the building in China where stolen western iPhones are stripped and soldYou’ll find a buyer for every single component inside the Feiyang Times. (FT $) 10 Amazon has started randomly refunding customers for old purchasesSome orders were placed as far back as 2018. (Bloomberg $) Quote of the day “Anybody who’s a computer scientist should not be retired right now. They should be working on AI.” —Google cofounder Sergey Brin says people with the right technical skills should copy him and quit being retired, TechCrunch reports. One more thing This fuel plant will use agricultural waste to combat climate changeA startup called Mote plans to build a new type of fuel-producing plant in California’s fertile Central Valley that would, if it works as hoped, continually capture and bury carbon dioxide.It’s among a growing number of efforts to commercialize a concept first proposed two decades ago as a means of combating climate change, known as bioenergy with carbon capture and sequestration, or BECCS.It’s an ambitious plan. However, there are serious challenges to doing BECCS affordably and in ways that reliably suck down significant levels of carbon dioxide. Read the full story. —James Temple We can still have nice things A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.) + These creepy little Labubu toys are everywhere. But why?+ Happy 25th birthday to one of London’s finest institutions, the Tate Modern gallery.+ Why the Mission Impossible film franchise just won’t die.+ Hummingbirds can fly backwards!? Wow.