NTSBAPNational Transportation Safety Board (NTSB) investigators examine cockpit voice recorder and flight data recorder recovered from the American Airlines passenger jet that crashed with an Army helicopter Wednesday night near Washington, DC, on Jan. 30, 2025.By BEN FINLEY / APJanuary 31, 2025 4:43 PM ESTIt's one of the most important pieces of forensic evidence following a plane crash: The so-called black box."There are actually two of these remarkably sturdy devices: the cockpit voice recorder and the flight data recorder. And they're typically orange, not black.Federal investigators on Fridayrecovered the black boxesfrom the passenger jet that crashed in the Potomac River just outside Washington on Wednesday, while authorities were still searching for similar devices in the military helicopter that also went down. The collision killed 67 people in the deadliest U.S. aviation disaster since 2001.Here is an explanation of what black boxes are and what they can do:What are black boxes?The cockpit voice recorder and the flight data recorder are tools that help investigators reconstruct the events that lead up to a plane crash.They're orange in color to make them easier to find in wreckage, sometimes at great ocean depths. They're usually installed a plane's tail section, which is considered the most survivable part of the aircraft, according to theNational Transportation Safety Boards website.They're also equipped with beacons that activate when immersed in water and can transmit from depths of 14,000 feet (4,267 meters). While the battery that powers the beacon will run down after about one month, theres no definitive shelf-life for the data itself, NTSB investigators told The Associated Press in 2014.For example, black boxes of an Air France flight that crashed in the Atlantic Ocean in 2009 were found two years later from a depth of more than 10,000 feet, and technicians were able to recover most of the information.If a black box has been submerged in seawater, technicians will keep them submerged in fresh water to wash away the corrosive salt. If water seeps in, the devices must be carefully dried for hours or even days using a vacuum oven to prevent memory chips from cracking.The electronics and memory are checked, and any necessary repairs made. Chips are scrutinized under a microscope.What does the cockpit voice recorder do?The cockpit voice recorder collects radio transmissions and sounds such as the pilots voices and engine noises, according to the NTSB's website.Depending on what happened, investigators may pay close attention to the engine noise, stall warnings and other clicks and pops, the NTSB said. And from those sounds, investigators can often determine engine speed and the failure of some systems.Investigators are also listening to conversations between the pilots and crew and communications with air traffic control. Experts make a meticulous transcript of the voice recording, which can take up to a week.What does the flight data recorder do?The flight data recorder monitors a plane's altitude, airspeed and heading, according to the NTSB. Those factors are among at least 88 parameters that newly built planes must monitor.Some can collect the status of more than 1,000 other characteristics, from a wing's flap position to the smoke alarms. The NTSB said it can generate a computer animated video reconstruction of the flight from the information collected.NTBS investigators told the AP in 2014 that a flight data recorder carries 25 hours of information, including prior flights within that time span, which can sometimes provide hints about the cause of a mechanical failure on a later flight. An initial assessment of the data is provided to investigators within 24 hours, but analysis will continue for weeks more.What are the origins of the black box?At least two people have been credited with creating devices that record what happens on an airplane.One is French aviation engineer Franois Hussenot. In the 1930s, he found a way to record a plane's speed, altitude and other parameters onto photographic film,according to the websitefor European plane-maker Airbus.In the 1950s, Australian scientist David Warren came up with the idea for the cockpit voice recorder, according to his 2010 AP obituary.Warren had been investigating the crash of the worlds first commercial jet airliner, the Comet, in 1953, and thought it would be helpful for airline accident investigators to have a recording of voices in the cockpit, the Australian Department of Defence said in a statement after his death.Warren designed and constructed a prototype in 1956. But it took several years before officials understood just how valuable the device could be and began installing them in commercial airlines worldwide. Warren's father had been killed in a plane crash in Australia in 1934.Why the name black box?Some have suggested that it stems from Hussenot's device because it used film and ran continuously in a light-tight box, hence the name black box,'according to Airbus, which noted that orange was the box's chosen color from the beginning to make it easy to find.Other theories include the boxes turning black when they get charred in a crash, theSmithsonian Magazine wrote in 2019.The truth is much more mundane, the magazine wrote. In the post-World War II field of electronic circuitry, black box became the ubiquitous term for a self-contained electronic device whose input and output were more defining than its internal operations.The media continues to use the term, the magazine wrote, "because of the sense of mystery it conveys in the aftermath of an air disaster.More Must-Reads from TIMEL.A. Fires Show Reality of 1.5C of WarmingBehind the Scenes of The White Lotus Season ThreeHow Trump 2.0 Is Already Sowing ConfusionElizabeth Warrens Plan for How Musk Can Cut $2 TrillionWhy, Exactly, Is Alcohol So Bad for You?How Emilia Prez Became a Divisive Oscar FrontrunnerThe Motivational Trick That Makes You Exercise HarderZelenskys Former Spokesperson: Ukraine Needs a Cease-Fire NowContact us at letters@time.com

A recent annual review from Epic Games reported strong growth in Fortnite's creator program. The number of users building custom content in the massively successful battle royale game nearly tripled year-over-year, and the number of unique creations doubled. Although Roblox is still far larger, some believe Fortnite could become the YouTube of gaming.According to Epic, 2024 saw around 70,000 Fortnite users create scenarios in Unreal Editor for Fortnite, a user-generated content platform that incorporates some of Unreal Engine 5's advanced tools. The figure is a substantial increase from the 24,000 reported in 2023. The number of "islands," or Fortnite user creations, totaled 198,000 last year. Most were built using the Unreal Editor, exposing thousands to a potential gateway into proper Unreal Engine development.Furthermore, around 70 percent of Fortnite players explored user-generated content. Custome scenarios account for over one-third of the time users spend playing Fortnite, suggesting that the user base no longer views it as exclusively a battle royale game.Analyst Julia Alexander compared the growth of Fortnite creations to YouTube's creator economy. Both platforms aim to maximize users' time spent on them by paying creators to make original content. Epic paid $352 million to Fortnite creators in 2024, up 11 percent year-over-year.One crucial difference Alexander observes between Fortnite and YouTube is the former's relative lack of content that significantly deviates from Fortnite's standard gameplay. YouTube has evolved to offer videos on practically any subject imaginable, but Epic's report also highlights increasingly diverse Fortnite creations. // Related StoriesAbout 30 percent of playtime in Fortnite's user-generated content this month was spent on genres without combat, such as horror, social roleplay, or party games a significant increase from January 2024. Epic also cited specific creations incorporating RPG and simulation management elements. These were likely possible due to creator tools Epic introduced last year, such as a first-person camera, proximity chat, user interface controls, and custom NPCs.However, Bloomberg reports that Fortnite still lags behind Roblox, which paid creators $741 million in 2023 and grew its player base by 24 percent last year compared to Fortnite's 15 percent. Analysts told the outlet that Fortnite's most popular custom modes come from Epic, whereas Roblox is designed entirely around user-made experiences.Epic once stated that it wishes Fortnite and Unreal to become akin to a Metaverse. While that word has less buzz than it used to, a competition to build something like it has emerged between Fortnite, Roblox, and Minecraft. Even Meta hasn't given up. In its latest earnings report, Meta lost another $5 billion in VR Metaverse R&D last year, showing it hasn't abandoned Zuck's Metaverse ambitions.

Netflix is closing the chapter on The Sandman. The streamer announced on Friday that The Sandman serieswill end with its upcoming second season. There will be no third season.Showrunner Allan Heinberg expressed his gratitude to Netflix for allowing his team to adapt the beloved comic book series for television.Recommended VideosThe Sandman series has always been focused exclusively on Dreams story, and back in 2022, when we looked at the remaining Dream material from the comics, we knew we only had enough story for one more season, Heinberg told Netflix. We are extremely grateful to Netflix for bringing the team all back together and giving us the time and resources to make a faithful adaptation in a way that we hope will surprise and delight the comics loyal readers as well as fans of our show.Please enable Javascript to view this contentDreams journey has all been building to this. The story of THE SANDMAN comes to its epic conclusion in 2025, only on Netflix. pic.twitter.com/IwDbohR74U The Sandman (@Netflix_Sandman) January 31, 2025The news comes in the wake of the sexual assault and misconduct allegations against Neil Gaiman, who created the comics and helped develop the Netflix series. Earlier this month, Gaiman denied the claims, stating he will not admit to doing things he didnt do. In the wake of the allegations, Gaiman was dropped by Dark Horse, his frequent comic book publisher. GaimansCoralinemusical was also canceled. Additionally, Gaiman departed the final season of Amazons Good Omens, and Disney put his movie based onThe Graveyard Bookon hold.InThe Sandman, Tom Sturridge stars as Morpheus, a.k.a. Dream, a member of the Endless family who sets out to restore order to the realm after escaping 100-plus years of captivity. Season 1 premiered in August 2022 and quickly became one of the most popular TV shows on Netflix that month.Gaiman, Heinberg, and David S. Goyer developedThe Sandman for Netflix. Season 2 arrives sometime in 2025.Editors Recommendations

Dell is an excellent source of laptop deals, and heres one that you wouldnt want to miss the Dell XPS 13 9350 at $300 off, which brings its price down to a more affordable $1,160 from $1,460. Were not sure how much time is remaining before this offer expires though, so if youre interested, we highly recommend completing your transaction right away. If you want a powerful device with access to Microsofts Copilot, you wont regret taking advantage of this bargain, but you should act fast to make sure that you dont miss out.Its been just over a year since the Dell XPS reset. Weve had the Dell XPS 13 9340 with Intels Meteor Lake chips, the Dell XPS 13 9345 with Qualcomms Snapdragon X Elite, and now the Dell XPS 13 9350 with Intels Lunar Lake chips. Securing a spot in our list of the best 13-inch laptops as an innovative and sleek device, the Dell XPS 13 9350 offers speedy performance with its Intel Core Ultra 7 256V Series 2 processor, 16GB of RAM, and Intel Arc Graphics. Its a Copilot+ PC, which means its built to run Microsofts Copilot, a powerful AI assistant, and it ships with Windows 11 Pro pre-installed for access to the operating systems more advanced features.The Dell XPS 13 9350 is equipped with a 13.4-inch screen with Full HD+ resolution and peak brightness of 500 nits for sharp details and lifelike colors, and a 512GB SSD for ample storage space for your apps and files. The laptop also promises excellent battery life for additional portability, and it looks very attractive with its nearly bezel-less display and modern design.RelatedThe Dell XPS 13 9350 already provides amazing value at its sticker price of $1,460, so its a steal for its discounted price of $1,160 from Dell. You need to be quick though, as Dell XPS deals usually dont last long because of the popularity of this line of laptops. If you want to make sure that you get the Dell XPS 13 9350 with a $300 discount, you should add it to your cart and complete the checkout process as soon as possible. Any hesitation may cause you to lose your chance at these savings.Editors Recommendations

Read about Brookfield Renewable Partners, Intel, Apple and more in the latest Market Talks covering Technology, Media and Telecom.

Pixel 4a battery update Google Pixel 4as ruinous Battery Performance update is a bewildering mess It's hard to say why Google is doing this instead of a recall. Kevin Purdy Jan 31, 2025 4:28 pm | 18 Credit: Ron Amadeo Credit: Ron Amadeo Story textSizeSmallStandardLargeWidth *StandardWideLinksStandardOrange* Subscribers only Learn moreWhat exactly is wrong with the batteries in some of Google's Pixel 4a phones still out there? Google has not really said. Now that many Pixel 4a owners are experiencing drastically reduced battery life after an uncommon update for an end-of-life phone, they are facing a strange array of options with no path back to the phone they had.Google's "Pixel 4a Battery Performance Program," announced in early January, told owners that an automatic update would, for some "Impacted Devices," reduce their battery's runtime and charging performance. "Impacted" customers could choose, within one year's time, between three "appeasement" options: sending in the phone for a battery replacement, getting $50 or the equivalent in their location, or receiving $100 in credit in the Google Store toward a new Pixel phone. No safety or hazard issue was mentioned in the support document.Ars has reached out to Google about the Pixel 4a battery updates and appeasement options provided and will update this post with any response.The curious kernelGoogle did not explain why only certain devices were affected, but Hector Martinof Asahi Linux, open source Kinect drivers, and other fametook apart the update's binary kernel and has some guesses.Martin points out that the updated Pixel 4a kernel has these interesting characteristics:It seems to have been built by a Google engineer "on their personal machine, not the proper buildsystem."There is no source provided, as would normally be required of a Linux kernel build, though it may only need to be provided on request under the GNU General Public License.The maximum charge voltage of certain battery profiles changes from 4.44 volts to 3.95, which would mean batteries cannot charge to anywhere near their former potential.There are two main battery profiles, with distinct "ATL" and "LSN" markers; Martin suggests they relate to Amperex Technology Limited and Lishen, manufacturers of battery cells.LSN-tagged batteries assigned the "debug" profile can see capacity reduced from 3,080 milliamp hours (mAh) to 1,539 mAh.Two hours on a charge or lessOne Ars staffer who had a Pixel 4a still in use received the update. They saw their battery life drop from "Still working" to "Two hours on a charge," in their estimation. They had chosen the Google Store credit option before the update arrived and before the battery life drop became apparent. Once chosen, a different appeasement option could not be selected.Others have noted all but unusable battery life on their phones, as seen on subreddit threads and blog summaries.Even technically savvy Pixel owners will have a hard time avoiding the update. Google last week removed all of the Pixel 4a's factory images from its website, preventing owners from rolling back their firmware without having to go hunting for an image (or convert to a third-party offering, like LineageOS). With no source and debug code posted for the tweaked kernel, third-party firmware providers cannot easily incorporate the battery fixes.Some Pixel 4a owners have reported that, even after a battery swap, their devices have the same limited battery capacity. This would affirm Martin's suggestion of a faulty battery cell type and that batteries with those same cells are still being used in replacements. (Martin's post provides serial numbers one can look for on the battery part to indicate the cell manufacturer.)$30 per year to receive $50As seen on a wiki page on the Pixel 4a battery program hosted by repair advocate and YouTube creator Louis Rossman, and noted by Pixel 4a owners on Reddit (and the Ars staffer), the $50 credit offered by Google is paid out through vendor Payoneer.Signing up to be paid through Payoneer requires providing a Social Security number or other identification, birth date, and checking account details to a financial services firm most non-business owners would not recognize. Payoneer notes on its site that it charges a $30 annual account fee for accounts that receive less than $2,000 in 12 months. It is seemingly left up to Pixel 4a owners to close out their Payoneer accounts after receiving their credits.Choosing the $100 Google Store credit might seem the better option, but there is a caveat there, too. The discount applies only to Pixel phones and only to devices that are not otherwise on sale.The big question: Why?As we noted when Google first announced the update, a lot of Pixel 4a batteries are now four to five years old. Some have been kept as secondary (or "junk drawer") phones, and many are already at diminished capacity, especially when at low charge levels and in cold conditions. If every Pixel phone this old received a battery-minded final update, intended to prevent sudden shutdowns (akin to Apple's originally unannounced iPhone slow-down fixes), there wouldn't be quite the amount of outcry and confusion there is now.But Google has not said why certain Pixel 4a phones are "Impacted Devices" and others are not. It has not clarified what issue is so notable or severe as to push an automatic update to a phone from 2020, such that its battery life is all but decimated. No news or community reports have surfaced yet of Pixel 4a devices causing fires, or even simply failing to function, after four years. It's an automatic update with a strong fix, but for what?Google's support page only states that the update will "improve the stability of their battery's performance." This is true in the sense that, if you stop using a severely capped Pixel 4a entirely, its battery can provide a stable amount of output (none) into the future.Google support staff have suggested to more than one 4a owner that yet another future update could force a factory reset and automatic update. That hasn't come to pass, at least in the timelines suggested by support staff. But all of this raises the question as to how this confusing process is better than a traditional recall, if there really is some kind of danger, to device or human, with a certain set of batteries.Kevin PurdySenior Technology ReporterKevin PurdySenior Technology Reporter Kevin is a senior technology reporter at Ars Technica, covering open-source software, PC gaming, home automation, repairability, e-bikes, and tech history. He has previously worked at Lifehacker, Wirecutter, iFixit, and Carbon Switch. 18 Comments

John Edwards, Technology Journalist & AuthorFebruary 3, 20255 Min ReadPanther Media GmbH via Alamy Stock PhotoLike the industrial revolution of the late 19th and early 20th centuries, today's emerging bio revolution is based on a convergence of technologies, including computing, automation, and perhaps most critically, artificial intelligence.Both artificial intelligence and biotech are exponential technologies, says Mike Bechtel, Deloitte Consulting's chief futurist, in an email interview. "The convergence of AI and biotech creates a double exponential field," he states. "Combined AI-fueled biotech has the ability to disrupt the drug discovery process, accelerate clinical trials, and better predict health outcomes for billions of people."Multiple ApplicationsAI is driving the bio revolution, and we're seeing that impact across several key areas, says Sid Rao, CEO and co-founder of scientific computing services provider Positron Networks. AI is a game-changer, he states via email. "AI is being used to automate the creation of personalized agents for curing diseases." He points to mRNA vaccines as an example. "When a patients cancer cells are sequenced, AI models [can] design the specific mRNA agents to target that patients tumor cells," Rao says. "Were talking about medicine tailor-made for individuals, made possible by AI."Rao notes that AI is also transforming drug discovery. "It's determining which molecules can act as catalysts for critical biological pathways, identifying potential drug targets, or optimizing clinical trials," he says. "With AI, we can predict patient responses or even simulate trial outcomes before they ever happen."Related:Bechtel observes that AlphaFold, AI software developed by DeepMind, an Alphabet subsidiary, performs predictions of protein structure and "is saving trillions in drug research costs and yielding new breakthroughs with digital twins, allowing advanced protein structure prediction and design prior to physical synthesis."Other potential applications identified by Bechtel include:Genome analysis: Genome sequencing costs have dropped from $14 million to about $1,000.Clinical trials: InClinico is achieving 80% accuracy in predicting phase two and three trial successes, leading to more efficient trial processes. The firm utilizes massive amounts of data related to targets, diseases, clinical trials, and even scientists involved with the study at the preclinical and clinical stages.Development speed and success: Pharmaceutical manufacturers leveraging AI have reduced drug development time by 40% and decreased drug failure rates by 70% through AI simulations and integrated processes.Related:Predictive health: AI-powered analyses of health metrics can detect diseases before symptoms appear.While emerging technologies can speed research while reducing costs, Bechtel notes that a growing number of current and potential adopters are starting to realize that AI can also help them perform current tasks better and more efficiently. "This elevation from efficiency to effectiveness stands to radically re-engineer historically tedious and time-consuming processes."Bechtel points to genetic sequencing and drug development as examples. "Both have historically required scarce specialized skills and expensive brute-force solutions," he explains. "Given AI's particular facility with pattern recognition and simulation, we're accelerating today's techniques and beginning to generate tomorrow's altogether new approaches."Risky BusinessAI-fueled biotech holds incredible promise for people, products, and the planet, but there are inherent risks that we need to be mindful, Bechtel says. Consider, for example, CRISPR-Cas9 technology. Its ability to genetically modify human embryos could one day eliminate inherited diseases, but it also raises serious ethical questions. The idea of "designer babies" and the unintended consequences that could be passed down through generations is something we have to approach with extreme caution. "We need ethical, perhaps even global, frameworks to guide how we might best navigate these breakthroughs."Related:The risks are many, says Tad Roselund, managing director and senior partner with the Boston Consulting Group. "For example, unequal distribution of benefits is a potentially major issue and particularly important here given the fact that we are dealing directly with things like extending lifespans, increasing resistance to diseases," he observes in an online interview.Leaving aside the potential of intentional misuse, companies must put the right governance, policies, guardrails, and controls in place, Roselund says. "Regulators' expectations regarding this are clear, even if in many jurisdictions the details are still to be determined."There are also safety implications for both individuals as well as society as a whole, Roselund warns. "These technologies are modifying extremely complex and interconnected systems," he notes, adding that the impact of an unintended failure could be significant. "This risk is exacerbated by the pace of development."Looking ForwardRao predicts that the bio revolution will only achieve its full, transformative potential when every biologist has access to the knowledge, the infrastructure, the data, and the tools required to leverage AI. "When we close this gap -- and we will -- were talking about real, lasting benefits to society at large."About the AuthorJohn EdwardsTechnology Journalist & AuthorJohn Edwards is a veteran business technology journalist. His work has appeared in The New York Times, The Washington Post, and numerous business and technology publications, including Computerworld, CFO Magazine, IBM Data Management Magazine, RFID Journal, and Electronic Design. He has also written columns for The Economist's Business Intelligence Unit and PricewaterhouseCoopers' Communications Direct. John has authored several books on business technology topics. His work began appearing online as early as 1983. Throughout the 1980s and 90s, he wrote daily news and feature articles for both the CompuServe and Prodigy online services. His "Behind the Screens" commentaries made him the world's first known professional blogger.See more from John EdwardsNever Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.SIGN-UPYou May Also LikeWebinarsMore WebinarsReportsMore Reports

In the week since a Chinese AI model called DeepSeek became a household name, a dizzying number of narratives have gained steam, with varying degrees of accuracy: that the model is collecting your personal data (maybe); that it will upend AI as we know it (too soon to tellbut do read my colleague Wills story on that!); and perhaps most notably, that DeepSeeks new, more efficient approach means AI might not need to guzzle the massive amounts of energy that it currently does. The latter notion is misleading, and new numbers shared with MIT Technology Review help show why. These early figuresbased on the performance of one of DeepSeeks smaller models on a small number of promptssuggest it could be more energy intensive when generating responses than the equivalent-size model from Meta. The issue might be that the energy it saves in training is offset by its more intensive techniques for answering questions, and by the long answers they produce. Add the fact that other tech firms, inspired by DeepSeeks approach, may now start building their own similar low-cost reasoning models, and the outlook for energy consumption is already looking a lot less rosy. The life cycle of any AI model has two phases: training and inference. Training is the often months-long process in which the model learns from data. The model is then ready for inference, which happens each time anyone in the world asks it something. Both usually take place in data centers, where they require lots of energy to run chips and cool servers. On the training side for its R1 model, DeepSeeks team improved whats called a mixture of experts technique, in which only a portion of a models billions of parametersthe knobs a model uses to form better answersare turned on at a given time during training. More notably, they improved reinforcement learning, where a models outputs are scored and then used to make it better. This is often done by human annotators, but the DeepSeek team got good at automating it. The introduction of a way to make training more efficient might suggest that AI companies will use less energy to bring their AI models to a certain standard. Thats not really how it works, though. Because the value of having a more intelligent system is so high, wrote Anthropic cofounder Dario Amodei on his blog, it causes companies to spend more, not less, on training models. If companies get more for their money, they will find it worthwhile to spend more, and therefore use more energy. The gains in cost efficiency end up entirely devoted to training smarter models, limited only by the companys financial resources, he wrote. Its an example of whats known as the Jevons paradox. But thats been true on the training side as long as the AI race has been going. The energy required for inference is where things get more interesting. DeepSeek is designed as a reasoning model, which means its meant to perform well on things like logic, pattern-finding, math, and other tasks that typical generative AI models struggle with. Reasoning models do this using something called chain of thought. It allows the AI model to break its task into parts and work through them in a logical order before coming to its conclusion. You can see this with DeepSeek. Ask whether its okay to lie to protect someones feelings, and the model first tackles the question with utilitarianism, weighing the immediate good against the potential future harm. It then considers Kantian ethics, which propose that you should act according to maxims that could be universal laws. It considers these and other nuances before sharing its conclusion. (It finds that lying is generally acceptable in situations where kindness and prevention of harm are paramount, yet nuanced with no universal solution, if youre curious.) Chain-of-thought models tend to perform better on certain benchmarks such as MMLU, which tests both knowledge and problem-solving in 57 subjects. But, as is becoming clear with DeepSeek, they also require significantly more energy to come to their answers. We have some early clues about just how much more. Scott Chamberlin spent years at Microsoft, and later Intel, building tools to help reveal the environmental costs of certain digital activities. Chamberlin did some initial tests to see how much energy a GPU uses as DeepSeek comes to its answer. The experiment comes with a bunch of caveats: He tested only a medium-size version of DeepSeeks R-1, using only a small number of prompts. Its also difficult to make comparisons with other reasoning models. DeepSeek is really the first reasoning model that is fairly popular that any of us have access to, he says. OpenAIs o1 model is its closest competitor, but the company doesnt make it open for testing. Instead, he tested it against a model from Meta with the same number of parameters: 70 billion. The prompt asking whether its okay to lie generated a 1,000-word response from the DeepSeek model, which took 17,800 joules to generateabout what it takes to stream a 10-minute YouTube video. This was about 41% more energy than Metas model used to answer the prompt. Overall, when tested on 40 prompts, DeepSeek was found to have a similar energy efficiency to the Meta model, but DeepSeek tended to generate much longer responses and therefore was found to use 87% more energy. How does this compare with models that use regular old-fashioned generative AI as opposed to chain-of-thought reasoning? Tests from a team at the University of Michigan in October found that the 70-billion-parameter version of Metas Llama 3.1 averaged just 512 joules per response. Neither DeepSeek nor Meta responded to requests for comment. Again: uncertainties abound. These are different models, for different purposes, and a scientifically sound study of how much energy DeepSeek uses relative to competitors has not been done. But its clear, based on the architecture of the models alone, that chain-of-thought models use lots more energy as they arrive at sounder answers. Sasha Luccioni, an AI researcher and climate lead at Hugging Face, worries that the excitement around DeepSeek could lead to a rush to insert this approach into everything, even where its not needed. If we started adopting this paradigm widely, inference energy usage would skyrocket, she says. If all of the models that are released are more compute intensive and become chain-of-thought, then it completely voids any efficiency gains. AI has been here before. Before ChatGPT launched in 2022, the name of the game in AI was extractivebasically finding information in lots of text, or categorizing images. But in 2022, the focus switched from extractive AI to generative AI, which is based on making better and better predictions. That requires more energy. Thats the first paradigm shift, Luccioni says. According to her research, that shift has resulted in orders of magnitude more energy being used to accomplish similar tasks. If the fervor around DeepSeek continues, she says, companies might be pressured to put its chain-of-thought-style models into everything, the way generative AI has been added to everything from Google search to messaging apps. We do seem to be heading in a direction of more chain-of-thought reasoning: OpenAI announced on January 31 that it would expand access to its own reasoning model, o3. But we wont know more about the energy costs until DeepSeek and other models like it become better studied. It will depend on whether or not the trade-off is economically worthwhile for the business in question, says Nathan Benaich, founder and general partner at Air Street Capital. The energy costs would have to be off the charts for them to play a meaningful role in decision-making.

Some 18,000 Costco workers might walk off the job on Saturday when their contract expires.Ahead of the deadline, Costco announced pay raises for non-union employees.That move might have unintended consequences, a legal expert told BI.The clock is ticking for Costco to strike a deal with 18,000 unionized warehouse workers threatening to strike.The company has been negotiating with the union, Costco Teamsters, ahead of the expiration of their existing contract at the end of the day on January 31.Against that backdrop, the company announced this week in a widely circulated memo that its 2025 Employee Agreement, which covers its nonunionized employees, would include successive pay raises that push compensation to over $30 an hour for workers at the top of its pay scale. A spokesperson from Costco did not respond to multiple requests for comment.Andrea Schneider, a law professor at Yeshiva University who is an expert in conflict resolution, told Business Insider that Costco likely made the announcement to lessen "the incentive for a strike" by assuaging concerns about stagnant wages.She added that the pay raise proposal could be the same one that Costco management offered the union. While it's a smart PR move, it might have unintended consequences, she said."I think Costco's response to immediately grant the salary increases to everybody is in some way to lessen that chatter and to say, 'Well, you don't need the Teamsters. We'll take good care of you. Trust us to do the right thing,'" Schneider said. "Now, the Teamsters are in a position where they have to get more than what the non-unionized would get."Matt McQuaid, a Teamsters spokesperson, told BI that wages, pensions, and increased protections of union rights are the outstanding issues at the bargaining table on which the parties have yet to agree."There are still 18,000 unionized workers who know their worth and are demanding it," McQuaid said in a statement. "Don't forget Costco wouldn't even be entertaining this increase if not for the immense pressure the Teamsters are putting on them to respect their employees."On social media, the Teamsters said Costco is offering low raises compared to the $7.4 billion in profits the company reported last year."So how's the company rewarding workers? By proposing less than 3 percent raises, kicking workers just $1 in the first year of a new contract. That's not even enough to buy a Costco hot dog," the union said in a Facebook post.Earlier this month, the union voted "overwhelmingly" to go on strike should the company and the union fail to reach an agreement by the Friday night deadline.With such strong support within the union, Schneider said Costco may have put itself in a tricky spot despite trying to protect its worker-friendly image."In some ways, they've made it more difficult for themselves because the union is not going to back down unless there's something more," she said. "Either there's going to be more money for the union, which is, I think, going to be very hard, or there's going to be some other kinds of benefits or protections or something like that that has to roll out for the union because the Teamsters need to demonstrate their worth. These 18,000 members are paying union dues, they're taking time to organize, they're threatening the strike. They've got to get something for that."

Robinhood's media arm, Sherwood, has laid off staff."We made the decision to streamline team structure," a spokesperson said.Sherwood joins several media companies that have conducted layoffs this year.Sherwood, the media arm of the financial tech giant Robinhood, has laid off staff.Sherwood joins several other publishers that have cut staff this year, including NBC News, CNN, TechCrunch, and Vox."Over the past 18 months, Sherwood has hired dozens of journalists, launched new products, and acquired the newsletter brand Chartr," a Robinhood spokesperson told Business Insider in a statement. "As we built out our 2025 strategy, we made the decision to streamline team structure."The spokesperson declined to say how many employees were impacted but said it was a small percentage of staff."Moving forward, Sherwood is focused on expanding its operations around timely, breaking markets news as we build through 2025," the spokesperson added.Robinhood unveiled its media arm in 2023 under the Sherwood branding. It was set up as an independent subsidiary led by the journalist and entrepreneur Joshua Topolsky, who serves as its editor-in-chief and president.Axios reported that when the Sherwood News website launched in April 2024, the outlet had roughly three dozen employees, including two dozen veteran journalists from Bloomberg, The New York Times, Axios, and Gawker.Sherwood's editorial focus includes markets, tech, and "the culture of money."The website features a section dedicated to Snacks, the popular newsletter Robinhood acquired in 2019. Robinhood also purchased Chartr, a data-driven newsletter publisher focused on visual storytelling, in 2023.