A ‘Sputnik' moment in the global AI race
Much remains uncertain about DeepSeek's LLM and its capabilities should not be overestimated — but its release nevertheless has sparked intense discussion about its superiority especially in terms of cost. DeepSeek claims that its model possesses reasoning abilities on par with or even superior to OpenAI's leading models, with training costs at less than one-tenth of OpenAI's — reportedly just $5.6 million — largely due to the use of NVIDIA's lower-cost H800 GPUs rather than the more powerful H200 or H100 models.
Tech giants like Meta and Google have spent billions of dollars on high-performance GPUs to develop cutting-edge AI models. However, DeepSeek's ability to produce a high-performance AI model at a significantly lower cost challenges the prevailing belief that computational power—determined by the number and quality of GPUs—is the primary driver of AI performance.
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Nikkei Asia
7 hours ago
- Nikkei Asia
OpenAI rolls out ChatGPT plan at about $4.60 in India to chase growth
India is OpenAI's second-largest market by user base after America. © Reuters August 19, 2025 13:35 JST (Reuters) -- ChatGPT maker OpenAI on Tuesday launched ChatGPT Go, a new India-only subscription plan priced at 399 rupees ($4.57) per month, its most affordable offering yet, as the company looks to deepen its presence in its second-largest market. Global companies often offer cheaper subscription plans for India's price-sensitive market, targeting the nearly 1 billion internet users in the world's most populous nation. The plan allows users to send up to 10 times more messages and generate 10 times more images compared to the free version, while also offering faster response times. Message limits increase with higher-tier subscription plans. ChatGPT Go is designed for Indians who want greater access to ChatGPT's advanced capabilities at a more affordable price, the Microsoft-backed startup said in a statement. The top-tier version of ChatGPT -- ChatGPT Pro -- is priced at 19,900 rupees per month in India, while ChatGPT Plus, its mid-range plan, costs 1,999 rupees a month. Earlier this year, CEO Sam Altman met with India's IT minister and discussed a plan to create a low-cost AI ecosystem. India is OpenAI's second-largest market by user base after the United States and may soon become the biggest, Altman said recently.
Yomiuri Shimbun
14 hours ago
- Yomiuri Shimbun
Making Cash off ‘AI Slop': The Surreal Video Business Taking over the Web
Luis Talavera, a 31-year-old loan officer in eastern Idaho, first went viral in June with an AI-generated video on TikTok in which a fake but lifelike old man talked about soiling himself. Within two weeks, he had used AI to pump out 91 more, mostly showing fake street interviews and jokes about fat people to an audience that has surged past 180,000 followers, some of whom comment to ask if the scenes are real. The low-effort, high-volume nature of AI videos has earned them the nickname 'AI slop,' and Talavera knows his videos aren't high art. But they earn him about $5,000 a month through TikTok's creator program, he said, so every night and weekend he spends hours churning them out. 'I've been on my couch holding my 3-month-old daughter, saying, 'Hey, ChatGPT, we're gonna create this script,'' he said. Nothing has transformed or polluted the creative landscape in the past few years quite like AI video, whose tools turn text commands into full-color footage that can look uncannily real. In the three years since ChatGPT's launch, AI videos have come to dominate the social web, copying and sometimes supplanting the human artists and videographers whose work helped train the systems in the first place. Their power has spawned a wild cottage industry of AI-video makers, enticed by the possibility of infinite creation for minimal work. Adele, a 20-year-old student in Florida who spoke on the condition that only her first name be used because she fears harassment, told The Washington Post she is taking a break from college to focus on making money from her AI-video accounts. Another creator in Arizona who went viral with an AI airport kangaroo said he made $15,000 in commissions in three months, speaking on the condition of anonymity out of concern over online harassment. But the flood of financially incentivized 'slop' has also given way to a strange new internet, where social media feeds overflow with unsettlingly lifelike imagery and even real videos can appear suspect. Some viral clips now barely rely on humans at all, with AI tools generating not just the imagery but also the ideas. 'I think of it more as a science than an art,' said one 25-year-old creator in Phoenix, who uses the online name Infinite Unreality and spoke on the condition of anonymity because he had received threats online. 'In reality, there's not a whole lot of creativity happening. And whatever creativity is happening is coming from the computer.' Some of the videos are otherworldly art pieces or cartoonishly goofy satires, carrying labels marking them as AI-made. But many others are deceptively realistic and styled as news reports, influencer posts or mean-spirited jokes, often in hopes they'll be shocking enough to grab attention – and from there, revenue. Built on tools from America's biggest tech giants, offered free or at low cost, the videos have touched off a kind of existential panic among the purveyors of traditional art, fueling anxiety that they could crowd out filmmakers, journalists and other creators for whom every scene takes money and time. 'As AI accelerates the production of content, human creativity will inevitably feel overwhelmed,' said Tony Sampson, a senior academic at the University of Essex who studies digital communication. AI videos don't try to compete on 'authenticity, aesthetic value or thought-provoking concepts,' he said. Instead, they're pumped out at industrial speed for maximum engagement, relying on viewers' shock and fascination to make them spread. The creators themselves say that AI videos are inevitable, regardless of their impact, and that they enjoy experimenting on AI's cutting edge. They are also eager to reap the rewards of mass attention: Juan Pablo Jiménez Domínguez, a 29-year-old creator known online as Pablo Prompt who works at a university in the Canary Islands, said he has used AI to create videos for ad campaigns and now makes enough that he 'could live entirely from this work.' 'A few months ago, we couldn't do half the things we can do now,' he said. The technology, he added, will help 'bring our ideas to life without the technical or financial blocks that used to hold us back.' 'A human being, just like you' The main benchmark for AI video is known as the Will Smith Eating Spaghetti test, and it works exactly how it sounds: A tool's progress is graded by how well it can make the actor look like he's chowing down. In 2023, the best versions looked muddy and deformed: Noodles oozed cartoonishly, eyes bugged out. This year's top performer, however, is practically undetectable as AI, save for one giveaway: The fake Smith makes crunching sounds, because the AI doesn't know how real spaghetti gets chewed. The quirk, in a Google-made tool called Veo 3, actually represents a major breakthrough for AI video: Unlike past tools, Veo 3 generates sound for every scene. And the progress continues rapidly: Google announced last month that the tool can now animate any photo into a lifelike eight-second clip. Every link of the AI-video supply chain has shown extraordinary progress over the past year, multiplying video makers' production power. A creator might, for instance, draft video ideas and dialogue with ChatGPT, generate images with Midjourney, compose realistic voices with ElevenLabs, and animate it all together with OpenAI's Sora, Meta's Movie Gen or a smaller upstart, such as Hailuo, Luma or Kling. In the late 2010s, amateurs used early, kludgy AI tools to splice women's faces into 'deepfake' pornography, soliciting money for individual requests. But the newer tools have made the process so simple that basically anyone can use them – as seen on Elon Musk's social network X, where users have prompted its AI tool Grok to create fake explicit videos of Taylor Swift. 'Five years ago, AI video was nonexistent to complete garbage. One year ago it was okay, not very usable, sort of just beginning,' said Mark Gadala-Maria, a co-founder at the AI tool Post Cheetah who tracks video trends. 'And today it's virtually indistinguishable from reality.' The shift has unleashed a barrage of AI video onto the web. In May, 4 of the 10 fastest-growing YouTube channels by subscribers trafficked in AI videos, an analysis in Sherwood News found, including Masters of Prophecy ('80s-style synthwave music videos) and Chick of Honor (nonsensical animal skits). Beyond video, there is AI music; one band, Velvet Sundown, had its AI-generated folk song 'Dust on the Wind' climb to the top of Spotify's Viral 50 charts, despite the fact that the band members don't actually exist. A single viral hit can spawn thousands of copycats, as with the videos of fake roller-coaster disasters, Bigfoot video diaries, jet-flying babies and Jiménez Domínguez's cats flipping off a diving board. To stand out, some creators have built AI-generated influencers with lives a viewer can follow along. 'Why does everybody think I'm AI? … I'm a human being, just like you guys,' says the AI woman in one since-removed TikTok video, which was watched more than 1 million times. The best-performing videos, Gadala-Maria said, have often relied on 'shock value,' such as racist and sexist jokes depicting Black women as primates, as first reported by Wired, or joking about what young 'AI gals gone wild' would do for cash. Others have ventured into dreamlike horror. One video showing a dog biting a woman's face off, revealing a salad, has more than 250 million views. The major social media platforms, scared of driving viewers away, have tried to crack down on slop accounts, using AI tools of their own to detect and flag videos they believe were synthetically made. YouTube last month said it would demonetize creators for 'inauthentic' and 'mass-produced' content. But the systems are imperfect, and the creators can easily spin up new accounts – or just push their AI tools to pump out videos similar to the banned ones, dodging attempts to snuff them out. 'Humans are attracted to things that are over the top,' Gadala-Maria said. 'And AI is really good at that.' 'Slop money' The typical AI creator's first dollar comes from the video platforms themselves, through the kinds of incentive programs that TikTok, YouTube and Instagram built to reward viral success. Adele, the Florida student, shared a screenshot from her TikTok account showing she had made $886 within four days from an AI-made video showing a fake influencer eating glass fruit. The 20-year-old said she had recently paused her psychology studies to focus on her entrepreneurial goals, including her 'AI Viral Club,' which offers video-making guides to roughly 70 subscribers paying $29 a month. 'I've seen a lot of my friends have a really hard time getting jobs, even with their degrees,' she said. 'This is the future.' Like Adele, many creators have worked to diversify beyond viral payouts, selling AI-tool courses and templates to aspiring creators eager to make their own. After an AI video showing two women eating a Korean-style 'mukbang' went viral, its creator began selling a $15 visual handbook on how others could copy its style. The creator, Jayla Bennett, who uses the account name ' said she is 26, works part-time in North Carolina and just started making AI videos this summer, seeing a chance at easy money. 'The trick is to get ahead of the curve and not be a part of the wave,' she said. Many creators also sell 'prompt drops,' listing the commands they gave the AI to make a certain scene, while others charge for custom-commissioned work. One creator said he is able to charge $200 to $300 for a five-second clip. Even bigger deals are being made. The prediction-gambling company Kalshi paid for a TV commercial during the NBA Finals featuring AI people, including a woman being battered by a hurricane, screaming their bets about current events. Jack Such, a Kalshi spokesman, said the video cost $2,000 in AI-prompting fees and went from idea to live in less than 72 hours, far quicker than a traditional studio could do. The creator, PJ Accetturo, said 'high-dopamine' AI videos would be 'the ad trend of 2025.' Angst over AI has roiled the traditional media for years, helping ignite the Hollywood strikes in 2023 and legal battles over artists' rights. In June, U.S. District Judge Vince Chhabria, ruling narrowly for Meta in a lawsuit brought by authors accusing the company of violating copyright law by training AI on their books, said AI would 'dramatically undermine the incentive for human beings to create things the old-fashioned way.' The technology has steadily inched its way into filmmaking nevertheless. Ted Sarandos, a co-chief of Netflix, said last month that the streamer had recently used AI video tools for the first time to help animate a building collapse for an Argentine sci-fi show, and that the move had been cheaper and '10 times faster' than a traditional special effects crew. Even for amateurs, AI video's ease of use has spawned a global business. Jiaru Tang, a researcher at the Queensland University of Technology who recently interviewed creators in China, said AI video has become one of the hottest new income opportunities there for workers in the internet's underbelly, who previously made money writing fake news articles or running spam accounts. Many university students, stay-at-home moms and the recently unemployed now see AI video as a kind of gig work, like driving an Uber. The average small creator she interviewed did their day jobs and then, at night, 'spent two to three hours making AI-slop money,' she said. A few she spoke with made $2,000 to $3,000 a month at it. 'They see their business as internet traffic, chasing really short-term trends, some of which are three to four days long,' said Patrik Wikstrom, a professor who oversaw Tang's research. 'They don't really care about if this is morally sound or if this is creative. They're chasing the traffic. They're chasing the next thing.' 'They just want to be entertained' For the AI-video creator Daryl Anselmo, this moment recalls a similarly massive shift known as the 'Demoscene,' an underground movement in the '90s built on computer nerds tinkering with real-time 3D graphics before 'Toy Story' and the Sony PlayStation made them mainstream. That era, too, stirred up apprehension among animators over the death of art. It also churned out a lot of slop. But the longtime video game artist has nevertheless gone all in on AI video, believing it offers a revolutionary new kind of artistic freedom. From his home in Vancouver, British Columbia, he describes himself as the creative director for a team of semiautonomous AI machines, each with its own task: visualizing ideas, generating images, animating scenes and stitching them together into experimental videos that are often skin-crawling and avant-garde. One morning, he takes photos of some leftover spinach and commands the system to make a spinach monster. It creates 10 different takes, and he chooses the leafiest and most menacing, telling the tools to examine the last video in the sequence and create the next most logical scene. All of the experimentation has cost him in the form of graphics-processing bills, which total thousands of dollars a month. But the videos have won him consulting work from companies eager to emulate what he calls his 'grimoire' of AI tools. And they've gained him attention on social media, where he often feels he must fight the urge to give people what they want: the creepiest videos, the most over-the-top. He can now churn out phantasmagoric scenes of hollow-eyed monsters at a speed and quality that would have once required a specialized team, but the pace of advancement slightly freaks him out. The 10 seconds of high-quality video it took him 10 minutes to create last year now takes just two minutes and, he expects, will soon take just a few seconds. At that speed, he said, creators could start really pushing the boundaries, rolling out hyper-personalized commercials and interactive videos a viewer could shape in real time, like a video game come to life. 'I don't know if we're prepared for the flood of generative media that's about to hit us,' he said. Even for those with less experience than Anselmo, this level of AI power has changed the industry. The creator in Phoenix, Infinite Unreality, started playing around with AI video while working in IT for his dad's company, hoping to spark a content-creator career with 'the most returns off minimal effort.' He made his first videos by taking viral clips on Instagram Reels and throwing them into Sora, asking the AI to transform them into something new. His first viral hit, which he described as 'some fat dude getting a massage,' gained 30 million views, but many more have followed. He now uses AI in every part of his eight-step workflow, from generating ideas to adding his logo in postproduction to discourage video thieves. His widely shared video of a seemingly real kangaroo at an airport gate, he said, took him 15 minutes, mainly because it was the AI tool's idea; he had asked it to spit out something that would go viral, and it did. 'I don't want to sit here and act like I'm this genius,' he said. 'I'm an entrepreneur.' But he has been unnerved by some of the responses to his more popular videos, including threats from viewers who say he's summoning something dark he can't control. His videos of lizard-headed babies – again, the AI tool's idea – have been especially unpopular. 'People are saying, 'This is disturbing.' But this is what you guys want to watch at the end of the day,' he said. It's not all easy money because many of the companies charge for individual AI-processing tasks; he budgets himself about $100 a day in AI fees, and a single 10-second rendering costs him about $7.50 to make. But he still expects to be made obsolete in short order, saying he believes that 'in a year from now, pretty much everything is going to be very easy for the average person' to make. His bigger fear, beyond how he'll pay his mortgage when that time comes, is a more existential one: about what all this limitless creation is doing to our brains. 'When you have every single form of media, every possible thing you can think of at your fingerprints at all times,' he said, 'is anything exciting anymore?' Gadala-Maria, in contrast, isn't worried. He expects AI one day will cement itself as the most powerful medium for human storytelling – not a meme or a novelty but an art form all its own. 'People generally don't care how their entertainment is created,' he said. 'They just want to be entertained.'
Japan Times
a day ago
- Japan Times
Foxconn to operate SoftBank's Stargate AI server site in Ohio
Hon Hai Precision Industry Co. will operate a U.S. factory owned by SoftBank Group Corp., setting up what's in the running to be the first manufacturing site in the Japanese company's $500 billion Stargate venture with OpenAI and Oracle Corp. SoftBank is acquiring Hon Hai's electric-vehicle plant in Ohio but the Taiwanese company will continue to run the complex after turning it into an AI server production plant, Hon Hai Chairman Young Liu said, confirming a report. SoftBank will supply manufacturing gear to the factory, and a joint venture between the two companies will make AI data center-related equipment, Liu said. SoftBank is scouting a number of potential data center sites to serve as a flagship for Stargate, weighing their access to water, power and telecom networks. Hon Hai's participation represents a boon for SoftBank founder Masayoshi Son's ambition to be at the center of surging investment in artificial intelligence hardware. Hon Hai — known also as Foxconn — assembles Apple iPhones and Nvidia servers. In January, when U.S. President Donald Trump first unveiled Stargate, Son stood beside him and promised to deploy $100 billion "immediately' into data centers, electricity generation and other infrastructure to support AI. But the Stargate rollout stalled after economic risks stemming from U.S. tariffs and the rise of cheaper AI alternatives such as DeepSeek made it harder to create the pricing models to secure project financing, it was reported in May. SoftBank has denied that financing has been an issue for Stargate.
