Chicago Slop Chicago Sun-Times prints summer reading list full of fake books Reading list in advertorial supplement contains 66% made up books by real authors. Benj Edwards – May 20, 2025 10:52 am | 42 Credit: Jorg Greuel via Getty Images Credit: Jorg Greuel via Getty Images Story text Size Small Standard Large Width * Standard Wide Links Standard Orange * Subscribers only   Learn more On Sunday, the Chicago Sun-Times published an advertorial summer reading list containing at least 10 fake books attributed to real authors, according to multiple reports on social media. The newspaper's uncredited "Summer reading list for 2025" supplement recommended titles including "Tidewater Dreams" by Isabel Allende and "The Last Algorithm" by Andy Weir—books that don't exist and were created out of thin air by an AI system. The creator of the list, Marco Buscaglia, confirmed to 404 Media that he used AI to generate the content. "I do use AI for background at times but always check out the material first. This time, I did not and I can't believe I missed it because it's so obvious. No excuses," Buscaglia said. "On me 100 percent and I'm completely embarrassed." A check by Ars Technica shows that only five of the fifteen recommended books in the list actually exist, with the remainder being fabricated titles falsely attributed to well-known authors. AI assistants such as ChatGPT are well-known for creating plausible-sounding errors known as confabulations, especially when lacking detailed information on a particular topic. The problem affects everything from AI search results to lawyers citing fake cases. On Tuesday morning, the Chicago Sun-Times addressed the controversy on Bluesky. "We are looking into how this made it into print as we speak," the official publication account wrote. "It is not editorial content and was not created by, or approved by, the Sun-Times newsroom. We value your trust in our reporting and take this very seriously. More info will be provided soon." In the supplement, the books listed by authors Isabel Allende, Andy Weir, Brit Bennett, Taylor Jenkins Reid, Min Jin Lee, Percival Everett, Delia Owens, Rumaan Alam, Rebecca Makkai, and Maggie O’Farrell are confabulated, while books listed by authors Françoise Sagan, Ray Bradbury, Jess Walter, André Aciman, and Ian McEwan are real. All of the authors are real people. Photo of the Chicago Sun-Times "Summer reading list for 2025" supplement. Credit: Rachel King / Bluesky Novelist Rachael King initially called attention to the error on Bluesky Tuesday morning. "The Chicago Sun-Times obviously gets ChatGPT to write a 'summer reads' feature almost entirely made up of real authors but completely fake books. What are we coming to?" King wrote. So far, community reaction to the list has been largely negative online, but others have expressed sympathy for the publication. Freelance journalist Joshua J. Friedman noted on Bluesky that the reading list was "part of a ~60-page summer supplement" published on May 18, suggesting it might be "transparent filler" possibly created by "the lone freelancer apparently saddled with producing it." The staffing connection The reading list appeared in a 64-page supplement called "Heat Index," which was a promotional section not specific to Chicago. Buscaglia told 404 Media the content was meant to be "generic and national" and would be inserted into newspapers around the country. "We never get a list of where things ran," he said. The publication error comes two months after the Chicago Sun-Times lost 20 percent of its staff through a buyout program. In March, the newspaper's nonprofit owner, Chicago Public Media, announced that 30 Sun-Times employees—including 23 from the newsroom—had accepted buyout offers amid financial struggles. A March report on the buyout in the Sun-Times described the staff reduction as "the most drastic the oft-imperiled Sun-Times has faced in several years." The departures included columnists, editorial writers, and editors with decades of experience. Melissa Bell, CEO of Chicago Public Media, stated at the time that the exits would save the company $4.2 million annually. The company offered buyouts as it prepared for an expected expiration of grant support at the end of 2026. Even with those pressures in the media, one Reddit user expressed disapproval of the apparent use of AI in the newspaper, even in a supplement that might not have been produced by staff. "As a subscriber, I am livid! What is the point of subscribing to a hard copy paper if they are just going to include AI slop too!?" wrote Reddit user xxxlovelit, who shared the reading list. "The Sun Times needs to answer for this, and there should be a reporter fired." This article was updated on May 20, 2025 at 11:02 AM to include information on Marco Buscaglia from 404 Media. Benj Edwards Senior AI Reporter Benj Edwards Senior AI Reporter Benj Edwards is Ars Technica's Senior AI Reporter and founder of the site's dedicated AI beat in 2022. He's also a tech historian with almost two decades of experience. In his free time, he writes and records music, collects vintage computers, and enjoys nature. He lives in Raleigh, NC. 42 Comments

When we set out to write a story on the best available estimates for AI’s energy and emissions burden, we knew there would be caveats and uncertainties to these numbers. But, we quickly discovered, the caveats are the story too.  This story is a part of MIT Technology Review’s series “Power Hungry: AI and our energy future,” on the energy demands and carbon costs of the artificial-intelligence revolution. Measuring the energy used by an AI model is not like evaluating a car’s fuel economy or an appliance’s energy rating. There’s no agreed-upon method or public database of values. There are no regulators who enforce standards, and consumers don’t get the chance to evaluate one model against another.  Despite the fact that billions of dollars are being poured into reshaping energy infrastructure around the needs of AI, no one has settled on a way to quantify AI’s energy usage. Worse, companies are generally unwilling to disclose their own piece of the puzzle. There are also limitations to estimating the emissions associated with that energy demand, because the grid hosts a complicated, ever-changing mix of energy sources.  It’s a big mess, basically. So, that said, here are the many variables, assumptions, and caveats that we used to calculate the consequences of an AI query. (You can see the full results of our investigation here.) Measuring the energy a model uses Companies like OpenAI, dealing in “closed-source” models, generally offer access to their  systems through an interface where you input a question and receive an answer. What happens in between—which data center in the world processes your request, the energy it takes to do so, and the carbon intensity of the energy sources used—remains a secret, knowable only to the companies. There are few incentives for them to release this information, and so far, most have not. That’s why, for our analysis, we looked at open-source models. They serve as a very imperfect proxy but the best one we have. (OpenAI, Microsoft, and Google declined to share specifics on how much energy their closed-source models use.)  The best resources for measuring the energy consumption of open-source AI models are AI Energy Score, ML.Energy, and MLPerf Power. The team behind ML.Energy assisted us with our text and image model calculations, and the team behind AI Energy Score helped with our video model calculations. Text models AI models use up energy in two phases: when they initially learn from vast amounts of data, called training, and when they respond to queries, called inference. When ChatGPT was launched a few years ago, training was the focus, as tech companies raced to keep up and build ever-bigger models. But now, inference is where the most energy is used. The most accurate way to understand how much energy an AI model uses in the inference stage is to directly measure the amount of electricity used by the server handling the request. Servers contain all sorts of components—powerful chips called GPUs that do the bulk of the computing, other chips called CPUs, fans to keep everything cool, and more. Researchers typically measure the amount of power the GPU draws and estimate the rest (more on this shortly).  To do this, we turned to PhD candidate Jae-Won Chung and associate professor Mosharaf Chowdhury at the University of Michigan, who lead the ML.Energy project. Once we collected figures for different models’ GPU energy use from their team, we had to estimate how much energy is used for other processes, like cooling. We examined research literature, including a 2024 paper from Microsoft, to understand how much of a server’s total energy demand GPUs are responsible for. It turns out to be about half. So we took the team’s GPU energy estimate and doubled it to get a sense of total energy demands.  The ML.Energy team uses a batch of 500 prompts from a larger dataset to test models. The hardware is kept the same throughout; the GPU is a popular Nvidia chip called the H100. We decided to focus on models of three sizes from the Meta Llama family: small (8 billion parameters), medium (70 billion), and large (405 billion). We also identified a selection of prompts to test. We compared these with the averages for the entire batch of 500 prompts.  Image models Stable Diffusion 3 from Stability AI is one of the most commonly used open-source image-generating models, so we made it our focus. Though we tested multiple sizes of the text-based Meta Llama model, we focused on one of the most popular sizes of Stable Diffusion 3, with 2 billion parameters.  The team uses a dataset of example prompts to test a model’s energy requirements. Though the energy used by large language models is determined partially by the prompt, this isn’t true for diffusion models. Diffusion models can be programmed to go through a prescribed number of “denoising steps” when they generate an image or video, with each step being an iteration of the algorithm that adds more detail to the image. For a given step count and model, all images generated have the same energy footprint. The more steps, the higher quality the end result—but the more energy used. Numbers of steps vary by model and application, but 25 is pretty common, and that’s what we used for our standard quality. For higher quality, we used 50 steps.  We mentioned that GPUs are usually responsible for about half of the energy demands of large language model requests. There is not sufficient research to know how this changes for diffusion models that generate images and videos. In the absence of a better estimate, and after consulting with researchers, we opted to stick with this 50% rule of thumb for images and videos too. Video models Chung and Chowdhury do test video models, but only ones that generate short, low-quality GIFs. We don’t think the videos these models produce mirror the fidelity of the AI-generated video that many people are used to seeing.  Instead, we turned to Sasha Luccioni, the AI and climate lead at Hugging Face, who directs the AI Energy Score project. She measures the energy used by the GPU during AI requests. We chose two versions of the CogVideoX model to test: an older, lower-quality version and a newer, higher-quality one.  We asked Luccioni to use her tool, called Code Carbon, to test both and measure the results of a batch of video prompts we selected, using the same hardware as our text and image tests to keep as many variables as possible the same. She reported the GPU energy demands, which we again doubled to estimate total energy demands.  Tracing where that energy comes from After we understand how much energy it takes to respond to a query, we can translate that into the total emissions impact. Doing so requires looking at the power grid from which data centers draw their electricity.  Nailing down the climate impact of the grid can be complicated, because it’s both interconnected and incredibly local. Imagine the grid as a system of connected canals and pools of water. Power plants add water to the canals, and electricity users, or loads, siphon it out. In the US, grid interconnections stretch all the way across the country. So, in a way, we’re all connected, but we can also break the grid up into its component pieces to get a sense for how energy sources vary across the country.  Understanding carbon intensity The key metric to understand here is called carbon intensity, which is basically a measure of how many grams of carbon dioxide pollution are released for every kilowatt-hour of electricity that’s produced.  To get carbon intensity figures, we reached out to Electricity Maps, a Danish startup company that gathers data on grids around the world. The team collects information from sources including governments and utilities and uses them to publish historical and real-time estimates of the carbon intensity of the grid. You can find more about their methodology here.  The company shared with us historical data from 2024, both for the entire US and for a few key balancing authorities (more on this in a moment). After discussions with Electricity Maps founder Olivier Corradi and other experts, we made a few decisions about which figures we would use in our calculations.  One way to measure carbon intensity is to simply look at all the power plants that are operating on the grid, add up the pollution they’re producing at the moment, and divide that total by the electricity they’re producing. But that doesn’t account for the emissions that are associated with building and tearing down power plants, which can be significant. So we chose to use carbon intensity figures that account for the whole life cycle of a power plant.  We also chose to use the consumption-based carbon intensity of energy rather than production-based. This figure accounts for imports and exports moving between different parts of the grid and best represents the electricity that’s being used, in real time, within a given region.  For most of the calculations you see in the story, we used the average carbon intensity for the US for 2024, according to Electricity Maps, which is 402.49 grams of carbon dioxide equivalent per kilowatt-hour.  Understanding balancing authorities While understanding the picture across the entire US can be helpful, the grid can look incredibly different in different locations.  One way we can break things up is by looking at balancing authorities. These are independent bodies responsible for grid balancing in a specific region. They operate mostly independently, though there’s a constant movement of electricity between them as well. There are 66 balancing authorities in the US, and we can calculate a carbon intensity for the part of the grid encompassed by a specific balancing authority. Electricity Maps provided carbon intensity figures for a few key balancing authorities, and we focused on several that play the largest roles in data center operations. ERCOT (which covers most of Texas) and PJM (a cluster of states on the East Coast, including Virginia, Pennsylvania, and New Jersey) are two of the regions with the largest burden of data centers, according to research from the Harvard School of Public Health.  We added CAISO (in California) because it covers the most populated state in the US. CAISO also manages a grid with a significant number of renewable energy sources, making it a good example of how carbon intensity can change drastically depending on the time of day. (In the middle of the day, solar tends to dominate, while natural gas plays a larger role overnight, for example.) One key caveat here is that we’re not entirely sure where companies tend to send individual AI inference requests. There are clusters of data centers in the regions we chose as examples, but when you use a tech giant’s AI model, your request could be handled by any number of data centers owned or contracted by the company. One reasonable approximation is location: It’s likely that the data center servicing a request is close to where it’s being made, so a request on the West Coast might be most likely to be routed to a data center on that side of the country.  Explaining what we found To better contextualize our calculations, we introduced a few comparisons people might be more familiar with than kilowatt-hours and grams of carbon dioxide. In a few places, we took the amount of electricity estimated to be used by a model and calculated how long that electricity would be able to power a standard microwave, as well as how far it might take someone on an e-bike.  In the case of the e-bike, we assumed an efficiency of 25 watt-hours per mile, which falls in the range of frequently cited efficiencies for a pedal-assisted bike. For the microwave, we assumed an 800-watt model, which falls within the average range in the US.  We also introduced a comparison to contextualize greenhouse gas emissions: miles driven in a gas-powered car. For this, we used data from the US Environmental Protection Agency, which puts the weighted average fuel economy of vehicles in the US in 2022 at 393 grams of carbon dioxide equivalent per mile.  Predicting how much energy AI will use in the future After measuring the energy demand of an individual query and the emissions it generated, it was time to estimate how all of this added up to national demand.  There are two ways to do this. In a bottom-up analysis, you estimate how many individual queries there are, calculate the energy demands of each, and add them up to determine the total. For a top-down look, you estimate how much energy all data centers are using by looking at larger trends.  Bottom-up is particularly difficult, because, once again, closed-source companies do not share such information and declined to talk specifics with us. While we can make some educated guesses to give us a picture of what might be happening right now, looking into the future is perhaps better served by taking a top-down approach. This data is scarce as well. The most important report was published in December by the Lawrence Berkeley National Laboratory, which is funded by the Department of Energy, and the report authors noted that it’s only the third such report released in the last 20 years. Academic climate and energy researchers we spoke with said it’s a major problem that AI is not considered its own economic sector for emissions measurements, and there aren’t rigorous reporting requirements. As a result, it’s difficult to track AI’s climate toll.  Still, we examined the report’s results, compared them with other findings and estimates, and consulted independent experts about the data. While much of the report was about data centers more broadly, we drew out data points that were specific to the future of AI.  Company goals We wanted to contrast these figures with the amounts of energy that AI companies themselves say they need. To do so, we collected reports by leading tech and AI companies about their plans for energy and data center expansions, as well as the dollar amounts they promised to invest. Where possible, we fact-checked the promises made in these claims. (Meta and Microsoft’s pledges to use more nuclear power, for example, would indeed reduce the carbon emissions of the companies, but it will take years, if not decades, for these additional nuclear plants to come online.)  Requests to companies We submitted requests to Microsoft, Google, and OpenAI to have data-driven conversations about their models’ energy demands for AI inference. None of the companies made executives or leadership available for on-the-record interviews about their energy usage. This story was supported by a grant from the Tarbell Center for AI Journalism.

When we set out to write a story on the best available estimates for AI’s energy and emissions burden, we knew there would be caveats and uncertainties to these numbers. But, we quickly discovered, the caveats are the story too.  This story is a part of MIT Technology Review’s series “Power Hungry: AI and our energy future,” on the energy demands and carbon costs of the artificial-intelligence revolution. Measuring the energy used by an AI model is not like evaluating a car’s fuel economy or an appliance’s energy rating. There’s no agreed-upon method or public database of values. There are no regulators who enforce standards, and consumers don’t get the chance to evaluate one model against another.  Despite the fact that billions of dollars are being poured into reshaping energy infrastructure around the needs of AI, no one has settled on a way to quantify AI’s energy usage. Worse, companies are generally unwilling to disclose their own piece of the puzzle. There are also limitations to estimating the emissions associated with that energy demand, because the grid hosts a complicated, ever-changing mix of energy sources.  It’s a big mess, basically. So, that said, here are the many variables, assumptions, and caveats that we used to calculate the consequences of an AI query. (You can see the full results of our investigation here.) Measuring the energy a model uses Companies like OpenAI, dealing in “closed-source” models, generally offer access to their  systems through an interface where you input a question and receive an answer. What happens in between—which data center in the world processes your request, the energy it takes to do so, and the carbon intensity of the energy sources used—remains a secret, knowable only to the companies. There are few incentives for them to release this information, and so far, most have not. That’s why, for our analysis, we looked at open-source models. They serve as a very imperfect proxy but the best one we have. (OpenAI, Microsoft, and Google declined to share specifics on how much energy their closed-source models use.)  The best resources for measuring the energy consumption of open-source AI models are AI Energy Score, ML.Energy, and MLPerf Power. The team behind ML.Energy assisted us with our text and image model calculations, and the team behind AI Energy Score helped with our video model calculations. Text models AI models use up energy in two phases: when they initially learn from vast amounts of data, called training, and when they respond to queries, called inference. When ChatGPT was launched a few years ago, training was the focus, as tech companies raced to keep up and build ever-bigger models. But now, inference is where the most energy is used. The most accurate way to understand how much energy an AI model uses in the inference stage is to directly measure the amount of electricity used by the server handling the request. Servers contain all sorts of components—powerful chips called GPUs that do the bulk of the computing, other chips called CPUs, fans to keep everything cool, and more. Researchers typically measure the amount of power the GPU draws and estimate the rest (more on this shortly).  To do this, we turned to PhD candidate Jae-Won Chung and associate professor Mosharaf Chowdhury at the University of Michigan, who lead the ML.Energy project. Once we collected figures for different models’ GPU energy use from their team, we had to estimate how much energy is used for other processes, like cooling. We examined research literature, including a 2024 paper from Microsoft, to understand how much of a server’s total energy demand GPUs are responsible for. It turns out to be about half. So we took the team’s GPU energy estimate and doubled it to get a sense of total energy demands.  The ML.Energy team uses a batch of 500 prompts from a larger dataset to test models. The hardware is kept the same throughout; the GPU is a popular Nvidia chip called the H100. We decided to focus on models of three sizes from the Meta Llama family: small (8 billion parameters), medium (70 billion), and large (405 billion). We also identified a selection of prompts to test. We compared these with the averages for the entire batch of 500 prompts.  Image models Stable Diffusion 3 from Stability AI is one of the most commonly used open-source image-generating models, so we made it our focus. Though we tested multiple sizes of the text-based Meta Llama model, we focused on one of the most popular sizes of Stable Diffusion 3, with 2 billion parameters.  The team uses a dataset of example prompts to test a model’s energy requirements. Though the energy used by large language models is determined partially by the prompt, this isn’t true for diffusion models. Diffusion models can be programmed to go through a prescribed number of “denoising steps” when they generate an image or video, with each step being an iteration of the algorithm that adds more detail to the image. For a given step count and model, all images generated have the same energy footprint. The more steps, the higher quality the end result—but the more energy used. Numbers of steps vary by model and application, but 25 is pretty common, and that’s what we used for our standard quality. For higher quality, we used 50 steps.  We mentioned that GPUs are usually responsible for about half of the energy demands of large language model requests. There is not sufficient research to know how this changes for diffusion models that generate images and videos. In the absence of a better estimate, and after consulting with researchers, we opted to stick with this 50% rule of thumb for images and videos too. Video models Chung and Chowdhury do test video models, but only ones that generate short, low-quality GIFs. We don’t think the videos these models produce mirror the fidelity of the AI-generated video that many people are used to seeing.  Instead, we turned to Sasha Luccioni, the AI and climate lead at Hugging Face, who directs the AI Energy Score project. She measures the energy used by the GPU during AI requests. We chose two versions of the CogVideoX model to test: an older, lower-quality version and a newer, higher-quality one.  We asked Luccioni to use her tool, called Code Carbon, to test both and measure the results of a batch of video prompts we selected, using the same hardware as our text and image tests to keep as many variables as possible the same. She reported the GPU energy demands, which we again doubled to estimate total energy demands.  Tracing where that energy comes from After we understand how much energy it takes to respond to a query, we can translate that into the total emissions impact. Doing so requires looking at the power grid from which data centers draw their electricity.  Nailing down the climate impact of the grid can be complicated, because it’s both interconnected and incredibly local. Imagine the grid as a system of connected canals and pools of water. Power plants add water to the canals, and electricity users, or loads, siphon it out. In the US, grid interconnections stretch all the way across the country. So, in a way, we’re all connected, but we can also break the grid up into its component pieces to get a sense for how energy sources vary across the country.  Understanding carbon intensity The key metric to understand here is called carbon intensity, which is basically a measure of how many grams of carbon dioxide pollution are released for every kilowatt-hour of electricity that’s produced.  To get carbon intensity figures, we reached out to Electricity Maps, a Danish startup company that gathers data on grids around the world. The team collects information from sources including governments and utilities and uses them to publish historical and real-time estimates of the carbon intensity of the grid. You can find more about their methodology here.  The company shared with us historical data from 2024, both for the entire US and for a few key balancing authorities (more on this in a moment). After discussions with Electricity Maps founder Olivier Corradi and other experts, we made a few decisions about which figures we would use in our calculations.  One way to measure carbon intensity is to simply look at all the power plants that are operating on the grid, add up the pollution they’re producing at the moment, and divide that total by the electricity they’re producing. But that doesn’t account for the emissions that are associated with building and tearing down power plants, which can be significant. So we chose to use carbon intensity figures that account for the whole life cycle of a power plant.  We also chose to use the consumption-based carbon intensity of energy rather than production-based. This figure accounts for imports and exports moving between different parts of the grid and best represents the electricity that’s being used, in real time, within a given region.  For most of the calculations you see in the story, we used the average carbon intensity for the US for 2024, according to Electricity Maps, which is 402.49 grams of carbon dioxide equivalent per kilowatt-hour.  Understanding balancing authorities While understanding the picture across the entire US can be helpful, the grid can look incredibly different in different locations.  One way we can break things up is by looking at balancing authorities. These are independent bodies responsible for grid balancing in a specific region. They operate mostly independently, though there’s a constant movement of electricity between them as well. There are 66 balancing authorities in the US, and we can calculate a carbon intensity for the part of the grid encompassed by a specific balancing authority. Electricity Maps provided carbon intensity figures for a few key balancing authorities, and we focused on several that play the largest roles in data center operations. ERCOT (which covers most of Texas) and PJM (a cluster of states on the East Coast, including Virginia, Pennsylvania, and New Jersey) are two of the regions with the largest burden of data centers, according to research from the Harvard School of Public Health.  We added CAISO (in California) because it covers the most populated state in the US. CAISO also manages a grid with a significant number of renewable energy sources, making it a good example of how carbon intensity can change drastically depending on the time of day. (In the middle of the day, solar tends to dominate, while natural gas plays a larger role overnight, for example.) One key caveat here is that we’re not entirely sure where companies tend to send individual AI inference requests. There are clusters of data centers in the regions we chose as examples, but when you use a tech giant’s AI model, your request could be handled by any number of data centers owned or contracted by the company. One reasonable approximation is location: It’s likely that the data center servicing a request is close to where it’s being made, so a request on the West Coast might be most likely to be routed to a data center on that side of the country.  Explaining what we found To better contextualize our calculations, we introduced a few comparisons people might be more familiar with than kilowatt-hours and grams of carbon dioxide. In a few places, we took the amount of electricity estimated to be used by a model and calculated how long that electricity would be able to power a standard microwave, as well as how far it might take someone on an e-bike.  In the case of the e-bike, we assumed an efficiency of 25 watt-hours per mile, which falls in the range of frequently cited efficiencies for a pedal-assisted bike. For the microwave, we assumed an 800-watt model, which falls within the average range in the US.  We also introduced a comparison to contextualize greenhouse gas emissions: miles driven in a gas-powered car. For this, we used data from the US Environmental Protection Agency, which puts the weighted average fuel economy of vehicles in the US in 2022 at 393 grams of carbon dioxide equivalent per mile.  Predicting how much energy AI will use in the future After measuring the energy demand of an individual query and the emissions it generated, it was time to estimate how all of this added up to national demand.  There are two ways to do this. In a bottom-up analysis, you estimate how many individual queries there are, calculate the energy demands of each, and add them up to determine the total. For a top-down look, you estimate how much energy all data centers are using by looking at larger trends.  Bottom-up is particularly difficult, because, once again, closed-source companies do not share such information and declined to talk specifics with us. While we can make some educated guesses to give us a picture of what might be happening right now, looking into the future is perhaps better served by taking a top-down approach. This data is scarce as well. The most important report was published in December by the Lawrence Berkeley National Laboratory, which is funded by the Department of Energy, and the report authors noted that it’s only the third such report released in the last 20 years. Academic climate and energy researchers we spoke with said it’s a major problem that AI is not considered its own economic sector for emissions measurements, and there aren’t rigorous reporting requirements. As a result, it’s difficult to track AI’s climate toll.  Still, we examined the report’s results, compared them with other findings and estimates, and consulted independent experts about the data. While much of the report was about data centers more broadly, we drew out data points that were specific to the future of AI.  Company goals We wanted to contrast these figures with the amounts of energy that AI companies themselves say they need. To do so, we collected reports by leading tech and AI companies about their plans for energy and data center expansions, as well as the dollar amounts they promised to invest. Where possible, we fact-checked the promises made in these claims. (Meta and Microsoft’s pledges to use more nuclear power, for example, would indeed reduce the carbon emissions of the companies, but it will take years, if not decades, for these additional nuclear plants to come online.)  Requests to companies We submitted requests to Microsoft, Google, and OpenAI to have data-driven conversations about their models’ energy demands for AI inference. None of the companies made executives or leadership available for on-the-record interviews about their energy usage. This story was supported by a grant from the Tarbell Center for AI Journalism.

In this version, it's the developer who collects the coins.

Pray I don't alter it any further Labor dispute erupts over AI-voiced Darth Vader in Fortnite SAG-AFTRA claims Epic didn't negotiate video game AI voice replacement terms. Benj Edwards – May 19, 2025 4:50 pm | 46 David Prowse as Darth Vader and Carrie Fisher as Princess Leia filming the original Star Wars. Credit: Sunset Boulevard/Corbis via Getty Images David Prowse as Darth Vader and Carrie Fisher as Princess Leia filming the original Star Wars. Credit: Sunset Boulevard/Corbis via Getty Images Story text Size Small Standard Large Width * Standard Wide Links Standard Orange * Subscribers only   Learn more On Monday, SAG-AFTRA filed an unfair labor practice charge with the National Labor Relations Board against Epic subsidiary Llama Productions for implementing an AI-generated Darth Vader voice in Fortnite on Friday without first notifying or bargaining with the union, as their contract requires. Llama Productions is the official signatory to SAG-AFTRA's collective bargaining agreement for Fortnite, making it legally responsible for adhering to the union's terms regarding the employment of voice actors and other performers. "We celebrate the right of our members and their estates to control the use of their digital replicas and welcome the use of new technologies," SAG-AFTRA stated in a news release. "However, we must protect our right to bargain terms and conditions around uses of voice that replace the work of our members, including those who previously did the work of matching Darth Vader's iconic rhythm and tone in video games." An official promo image for Darth Vader in Fortnite. Credit: Disney / Starwars.com The union's complaint comes just days after the feature sparked a separate controversy when players discovered that they could manipulate the AI into using profanity and inappropriate language until Epic quickly implemented a fix. The AI-controlled in-game character uses Google's Gemini 2.0 to generate dialogue and ElevenLabs' Flash v2.5 AI model trained on the voice of the late James Earl Jones to speak real-time responses to player questions. For voice actors who previously portrayed Darth Vader in video games, the Fortnite feature starkly illustrates how AI voice synthesis could reshape their profession. While James Earl Jones created the iconic voice for films, at least 54 voice actors have performed as Vader in various media games over the years when Jones wasn't available—work that could vanish if AI replicas become the industry standard. The union strikes back SAG-AFTRA's labor complaint (which can be read online here) doesn't focus on the AI feature's technical problems or on permission from the Jones estate, which explicitly authorized the use of a synthesized version of his voice for the character in Fortnite. The late actor, who died in 2024, had signed over his Darth Vader voice rights before his death. Instead, the union's grievance centers on labor rights and collective bargaining. In the NLRB filing, SAG-AFTRA alleges that Llama Productions "failed and refused to bargain in good faith with the union by making unilateral changes to terms and conditions of employment, without providing notice to the union or the opportunity to bargain, by utilizing AI-generated voices to replace bargaining unit work on the Interactive Program Fortnite." The action comes amid SAG-AFTRA's ongoing interactive media strike, which began in July 2024 after negotiations with video game producers stalled primarily over AI protections. The strike continues, with more than 100 games signing interim agreements, while others, including those from major publishers like Epic, remain in dispute. Benj Edwards Senior AI Reporter Benj Edwards Senior AI Reporter Benj Edwards is Ars Technica's Senior AI Reporter and founder of the site's dedicated AI beat in 2022. He's also a tech historian with almost two decades of experience. In his free time, he writes and records music, collects vintage computers, and enjoys nature. He lives in Raleigh, NC. 46 Comments

Leading healthcare organizations across the globe are using agentic AI, robotics and digital twins of medical environments to enhance surgical precision, boost workflow efficiency, improve medical diagnoses and more. Physical AI and humanoid robots in hospitals have the potential to automate routine tasks, assist with patient care and address workforce shortages. This is especially crucial in places where challenges to optimal healthcare services are paramount. Such challenges include hospital overcrowding, an aging population, rising healthcare costs and a shortage of medical professionals, all of which are affecting Taiwan, as well as many other regions and countries. At the COMPUTEX trade show in Taipei, NVIDIA today showcased how leading Taiwan medical centers are collaborating with top system builders to integrate smart hospital technologies and other AI-powered healthcare solutions that can help reduce these issues and save millions of lives. Cathay General Hospital, Chang Gung Memorial Hospital (CGMH), National Taiwan University Hospital (NTUH) and Taichung Veterans General Hospital (TCVGH) are among the top centers in the region pioneering healthcare AI innovation. Deployed in collaboration with leading system builders such as Advantech, Onyx, Foxconn and YUAN, these solutions tap into NVIDIA’s agentic AI and robotics technologies, including the NVIDIA Holoscan and IGX platforms, NVIDIA Jetson for embedded computing and NVIDIA Omniverse for simulating virtual worlds with OpenUSD. CGMH Boosts AI-Powered Medical Imaging With an average of 8.2 million outpatient visits and 2.4 million hospitalizations a year,  CGMH estimates that a third of the Taiwanese population has sought treatment at its vast network of hospitals in Taipei and seven other cities. The organization is pioneering smart hospital innovation by enhancing surgical precision and workflow efficiency through advanced, AI-powered colonoscopy workflow solutions, developed in collaboration with Advantech and based on the NVIDIA Holoscan platform, which includes the Holoscan SDK and the Holoscan Sensor Bridge running on NVIDIA IGX. NVIDIA Holoscan is a real-time sensor processing platform for edge AI compute, while NVIDIA IGX offers enterprise-ready, industrial edge AI purpose-built for medical environments. Using these platforms, CGMH is accelerating AI integration in its colonoscopy diagnostics procedures. Deployed in gastrointestinal consultation rooms, the AI-powered tool collects colonoscopy streams to train a customized model built on Holoscan and provides real-time colonic polyps identification and classification. Colonoscopy tools at CGMH. Image courtesy of CGMH. CGMH’s AI infrastructure — comprising NVIDIA accelerated computing, NVIDIA DGX systems, the MONAI framework, NVIDIA TensorRT-LLM open-source library, NVIDIA Dynamo inference framework, and the NVIDIA NeMo and Clara platforms — enables accelerated research and development across the organization. CGMH serves nearly 50 AI agent models that daily help the hospital analyze medical imaging, improving diagnostic accuracy, throughput and real-time inference at scale. For example, NVIDIA Triton-powered AI sped newborn examination record processing by 10x. Cathay General Hospital Improves Diagnostics With AI Cathay General Hospital, a Taipei-based healthcare center that provides hospital management and medical services, has collaborated with National Taiwan University Hospital, medical computer manufacturer Onyx and software provider aetherAI to develop an AI-assisted colonoscopy system that highlights lesions, detects hard-to-spot polyps and issues alerts to help physicians with diagnoses. Polyp detection during colonoscopy. Image courtesy of aetherAI and Onyx. Powered by a compact, plug-and-play AI BOX device — built with the NVIDIA Jetson AGX Xavier module — the AI system is trained on over 400,000 high-quality, physician-annotated images collected from patients with diverse and severe lesions over four years. The system can achieve up to 95.8% accuracy and sensitivity, and studies have shown that it can improve adenoma detection rates by up to 30%. These enhancements assist physicians in reducing diagnostic errors and making more informed treatment decisions, ultimately contributing to improved patient outcomes. NTUH Detects Liver Tumors, Cardiovascular Risks With AI In the 100+ years since its founding, NTUH has nurtured countless professionals in medicine and is renowned for its trusted clinical care. The national teaching hospital is now adopting AI imaging to more quickly, accurately diagnose patients. NTUH’s HeaortaNet model, trained on more than 70,000 axial images from 200 patients, automates CT scan segmentation of the heart, including the aorta and other arteries, in 3D, enabling rapid analysis of risks for cardiovascular disease. The model, which achieves high segmentation accuracy for the pericardium and aorta, significantly reduced data processing time per case from an hour to about 0.4 seconds. In addition, NTUH collaborated with the Good Liver Foundation and system builder YUAN to develop a diagnostic-assistance system for liver cancer detection during ultrasounds. It taps into an NVIDIA Jetson Orin NX module and a deep learning model trained on more than 5,000 annotated ultrasound images to identify malignant and benign liver tumors in real time. YUAN and NTUH’s liver cancer detection system turns an ultrasound device into an AI-assisted diagnostic tool. Image courtesy of YUAN. NVIDIA DeepStream and TensorRT SDKs accelerate the system’s deep learning model, ultimately helping clinicians detect tumors earlier and more reliably. In addition, NTUH is using NVIDIA DGX to train AI models for its system that detects pancreatic cancer from CT scans. TCVGH Streamlines Multimodal Imaging and Clinical Documentation Workflows With AI  Taichung Veterans General Hospital (TCVGH), a medical center and a teaching hospital administered by the Veterans Affairs Council in Taipei, has partnered with Foxconn to build physical and digital robots to augment staffing, improving clinician productivity and patient experiences. Foxconn developed an AI system that can analyze medical images and spot signs of breast cancer earlier than traditional methods, using NVIDIA Hopper GPUs, NVIDIA DGX systems and the MONAI framework. By tapping into clinical data and multimodal AI imaging, the system creates 3D virtual breast models, quickly highlighting areas of concern in scans to help radiologists make faster, more confident decisions. Foxconn is also working with TCVGH to build smart hospital solutions like the AI nursing collaborative robot Nurabot and tapping into NVIDIA Omniverse to create real-time digital twins of hospital environments, including nursing stations, patient wards and corridors. These digital replicas serve as high-fidelity simulations where Jetson-powered service robots can be trained to autonomously deliver medical supplies throughout the hospital, ultimately improving care efficiency. AI nursing collaborative robot Nurabot. Image courtesy of Foxconn. In addition, TCVGH has developed and deployed its Co-Healer system, which integrates the Taiwanese native large language model TAIDE-LX-7B to streamline clinical documentation processes with agentic AI. Co-Healer, built on the NVIDIA Jetson Xavier NX module, processes and helps summarize medical documents — such as nursing progress notes and health education materials — and supports medical exam preparation by providing students with instant access to nursing guidelines and patient-specific protocols for clinical procedures and diagnostic tests. This helps healthcare workers alleviate burnout while giving patients a clearer understanding of their diagnoses. Learn more about the latest AI advancements in healthcare at NVIDIA GTC Taipei, running May 21-22 at COMPUTEX.

Key Takeaways France is the first country to have endorsed the UN Open Source Principles, which now takes the total endorsements to 36. The French government recently released La Suite numérique, which is an open-source collection of collaboration tools for French government officials. French AI startup Mistral AI is also developing open-source AI models that can beat DeepSeek. France has become the first country to endorse the UN Open Source Principles  – an initiative adopted by the UN Chief Executive Board’s Digital Technology Network (DTN) to support the development of open-source technology within the UN. In addition to France, the UN Open Source Principles added 19 more endorsers to its list, taking the total to 36. The initiative has laid down eight principles for the development of open-source projects. This includes making open source the default approach for project development. Moreover, there’s also a focus on building interoperable projects to boost reusability and enable diverse and inclusive contributions. The biggest benefit of going open source is transparency. The code for any program will be accessible to the public, ensuring anyone can inspect and improve it. It also establishes trust between the developers and users.  For instance, if you’re using an open-source health tracking app, you can view the exact details it collects, including personal information. This ensures there are no back doors and users are not left in the dark.  It’s worth noting that the European Commission has also published an Open Source Software Strategy 2020-2023 to encourage and leverage open source collaboration. Like the UN, the EC, too, focused on sharing and reusing software tech along with spreading low-cost benefits to society. French Government Supporting Open Source The French government has developed the La Suite numérique – an open-source digital collaboration platform offering tools that can replace the likes of Microsoft 365 in corporate settings.  For instance, its Visio tool can be used for video conferencing. It doesn’t impose any time restrictions on meetings, irrespective of the group size – a perfect alternative to MS Teams. Alternatively, you can use the instant messaging app Tchap, which offers a single communication channel for your entire team. Install the Docs app from La Suite numérique if you want to work with documents. This one comes with features like media imports and offline mode. Moreover, several users can work simultaneously, meaning real-time collaboration is as simple as any mainstream docs app.  La Suite numérique also has a cloud storage app called Drive, a webmail service called Messagerie, and a secure file transfer service called France Transfert. Several other tools are also under development. Currently, La Suite numérique is only available to the French public administration and for official government use. However, it’s important to ponder the kind of affordability benefits tools like this can bring. Businesses can end up saving the cost of licensing each individual app for hundreds of thousands of devices. France’s Open Source AI Another open-source initiative from France is Mistral AI. This AI startup is working on open-source AI models that are powerful enough to outperform DeepSeek. CEO Arthur Mensch said that the company’s core focus is on open-source technology, which can help companies ‘build on top of one another.’ This is an entirely different approach to OpenAI. Despite being named after open source technology, the Altman-driven company has kept its advanced models, such as ChatGPT, GPT-4, and DALL-E, as closed source projects. Imagine open-source technology as the peak form of community building. So, if someone wants to build something similar to or better than Mistral AI, they can view and take inspiration from its code, which can give them a head start and help them build something better. However, if you want to build something like ChatGPT, you’ll have no resources and will need to start from scratch. This is precisely what the United Nations wants to change, and open-source technology aligns well with its objectives of transparency and inclusion. Open-source projects allow smaller players who do not have access to endless resources to contribute to the overall development of critical AI projects. The French government seems to have caught the cue. With in-house open-source tech development, open-source AI startups, and endorsement of the UN Open Source Principles, France now stands at the forefront of the open-source movement. Anything Can Be Open Source It’s worth noting that open-source technology isn’t just limited to AI or a select few industries. Any service offered on the internet can go open source, which is evident by the kind of support the UN open source principles have received. For example, one of the endorsees is F-Droid, which is an open-source alternative to the Google Play Store. Every app listed on F-Droid is open source and completely free. Plus, it doesn’t use any hidden analytics or ads to gather user data, making it a much safer and more transparent app marketplace than what we commonly use. Image Credit – Caschys Blog The WordPress Foundation has also supported the initiative. It aims to build a stable content management system (CMS) for web publishing. The Apereo Foundation, a nonprofit built with the aim of helping open-source advancements in higher education, has also endorsed the UN Open Source Principles. Currently, Apereo has over 25 open-source projects focused on privacy and decentralization. As you can see, businesses in different sectors have come up to support the open-source initiative of the United Nations. However, open-source technology’s biggest ‘flaw’ is its anti-capitalistic nature. This simply means that for-profit organizations would not actively support the development of open-source tech. The reason is pretty apparent: why would you reveal the secret ingredient of your best-selling dish? Open source lowers costs and entry barriers in any industry, which are the two things capitalism survives on. So, as unfortunate as it is, we might never see a fully open-source approach from tech giants like Google, Microsoft, or OpenAI. However, with the development of open-source projects supported by various governments, the user base might gradually shift from traditional tech producers. That said, there’s still a long way to go before open-source products become mainstream. Krishi is a seasoned tech journalist with over four years of experience writing about PC hardware, consumer technology, and artificial intelligence.  Clarity and accessibility are at the core of Krishi’s writing style. He believes technology writing should empower readers—not confuse them—and he’s committed to ensuring his content is always easy to understand without sacrificing accuracy or depth. Over the years, Krishi has contributed to some of the most reputable names in the industry, including Techopedia, TechRadar, and Tom’s Guide. A man of many talents, Krishi has also proven his mettle as a crypto writer, tackling complex topics with both ease and zeal. His work spans various formats—from in-depth explainers and news coverage to feature pieces and buying guides.  Behind the scenes, Krishi operates from a dual-monitor setup (including a 29-inch LG UltraWide) that’s always buzzing with news feeds, technical documentation, and research notes, as well as the occasional gaming sessions that keep him fresh.  Krishi thrives on staying current, always ready to dive into the latest announcements, industry shifts, and their far-reaching impacts.  When he's not deep into research on the latest PC hardware news, Krishi would love to chat with you about day trading and the financial markets—oh! And cricket, as well. View all articles by Krishi Chowdhary Our editorial process The Tech Report editorial policy is centered on providing helpful, accurate content that offers real value to our readers. We only work with experienced writers who have specific knowledge in the topics they cover, including latest developments in technology, online privacy, cryptocurrencies, software, and more. Our editorial policy ensures that each topic is researched and curated by our in-house editors. We maintain rigorous journalistic standards, and every article is 100% written by real authors.

‘Robo Catcher Zero’ Collects Cute Mechanical Critters Posted on May 19, 2025May 19, 2025 by Joel Couture Robo Catcher Zero takes players to a district filled with broken machines and hidden robot buddies for you to locate and awaken. Finder’s been tasked with wrangling up friendly robots that have tucked themselves into all sorts of places throughout the Amp220Ω district. The reason for why these machines are all so shy isn’t made clear in the current demo of the game, but that’s fine. I don’t need a lot of good reasons to find cute metal buddies besides the fact that they’re adorable. And this tiny slice of demo has quite a few cute robots to scrounge up. Finding the robots takes more than just finding their hiding spots. The robots don’t seem too keen on being discovered, so you have to play a short minigame when you find one. A small ball-like pod will roll around the screen and you will need to catch it with your mouse (looking like a catcher’s mitt) a few times in order to collect your new pal. It’s not especially challenging most of the time (other than the fact that some of the pods move around in a hurry), but it does add a tiny bit of action to gathering the robots. And once you have a few, you can have them follow you around as you explore the rest of the world. Robo Catcher Zero seems to be focused on creating a relaxed experience in exploring its decayed mechanical landscapes, simply having you putter around and find more robots. There is a story that appears to be taking shape as you find more machines, though, the world humming back to life as your party grows. I’m curious to see where the full version will take this story and why all of these robots started hiding. Robo Catcher Zero is currently in development, but in the meantime, you can grab a demo or add it to your wishlist on Steam. About The Author

MotoGP racing, the pinnacle of motorbike competition, requires precision and skill–and some scraped elbows and knees.   Image: Ducati/Lenovo Get the Popular Science daily newsletter💡 Breakthroughs, discoveries, and DIY tips sent every weekday. MotoGP is high-speed, high-tech motorcycle racing. The fastest riders in the world compete on specialized, purpose-built motorcycles from companies like Ducati, Honda, Yamaha on the world stage in this series, which is considered the most prestigious in the game. Riders reach incredible speeds on their machines up to 220 miles per hour, and races can go 350 turns with gravity-defying leaning that scrapes elbows and knees. This Grand Prix is for the toughest of the tough on the moto circuit.  Behind the riders and the machinery is a boatload of data and technology; we’re talking massive amounts of data in numbers that make your head spin. As tech company Lenovo puts it, when it comes to MotoGP–a sport in which every millisecond counts–data can mean the difference between a first and last place finish.  Lenovo and Italian motorbike manufacturer Ducati work hand in hand to create a champion racing team with information gathered by 50-plus sensors that track tire rotation, speed, grip, acceleration, and more. Motorcycle racing, at the highest echelon, has evolved way beyond just the machinery or even the skill of the rider. Without the technology, any team would be left in the dust.  The data journey In the past, the only way to evaluate the outer limits of what a motorbike could endure was to disassemble a crashed bike. Today, a multi-pronged approach to data collection takes that process up several notches. Every time a bike enters the pit garage, an in-person team and a crew from the Ducati Lenovo Remote Garage at Ducati’s headquarters in Bologna, Italy analyzes performance.   Via 50 sensors on each bike, Ducati engineers collect tons of data they can parse and assess. Image: Ducati/Lenovo No matter where the team is racing, data streams in from the motorbike sensors. Off-site engineers analyze information from the track, making tiny tweaks to the bike and its operation. That way, every qualifying lap may be reviewed and fine tuned for the best competitive outcomes.  “We debrief after each session, aligning with the rider on the sections and corners in which we think that he can improve his performance,” says David Attisano, the data and performance analysis manager for Ducati Racing. “We review with the rider his maneuvers and what we can expect to improve and how. For example, if the data shows that the rider can brake two meters later than he is, that’s a huge distance in this race.”  For the MotoGP series, Ducati has three different teams with a total of six riders. Each Ducati rider receives the data not just about his own ride but for all the other riders, so they can track what the other five are doing. That’s a giant amount of data. “We can push each other to the limit because we can see what the other riders are doing and comparing, then we can highlight and suggest some changes in their riding style,” Attisano explains.  Through its partnership with technology company Lenovo, Ducati collects and analyzes data to push its bikes to the limit and win races. Image: Lenovo/Ducati Per official MotoGP rules, the sensors can’t transmit the data while the bikes are in motion. As soon as the bike gets to the garage, the engineers download the data. From this point, the data analysis process begins. At this point, the partnership is especially critical as the servers are developed by Lenovo specifically for Ducati. On top of that, there are data signals created with a custom algorithm and simulating based on engineering knowledge about the bikes. The team can even create a detailed “digital twin” of the course that maps each twist and turn, using 200GB of data per circuit. “If the rider tells us that the traction control is not working properly, our electronics engineers can adjust the ECU,” says Attisano. “In our competition, the human factor is very important. We rely on the data and the humans do the last mile to gain the result and to achieve victory.”  Sure, data clearly plays a big part, but human interpretation is still a critical aspect to translate the results. Even those who know nothing of motorcycles can create algorithms or artificial intelligence to offer some hints or tips, but without input from the team, the data has no context. The technology, Attisano emphasizes, is intended to augment the capabilities of human beings, not to replace them.  Italian motorbike manufacturer Ducati makes high-speed, high-tech machines. Image: Ducati/Lenovo “The behavior of the bike is very complicated–it’s very difficult to understand if you have not studied it at university,” he says. “In our experience, the exchange of knowledge between engineers and data scientists is very important.”  Ducati’s data collection on the competitive side feeds into the safety and reliability for all riders, including those who like to take their motorbike for a spin on any given Sunday; not just competitive situations. Every time the bike emerges from the pit at a MotoGP race, the engineers strive to be sure it’s in good form and safe for the rider. Every improvement via the data collected from 50 tiny sensors on each factory-made racing bike leads to a better ride for casual street riders, too.