Tag: Generative AI

  • HoYoverse Invests $14.6B in AI Before PGC Barcelona

    HoYoverse Invests $14.6B in AI Before PGC Barcelona

    Key Takeaway

    – HoYoverse is investing $14.6 billion over three years to build proprietary generative AI infrastructure for live-service games.
    – The investment focuses on vertical integration—developing custom GPU clusters and model-training frameworks instead of licensing third-party LLMs.
    – First public test of the pipeline will be in *Petit Planet*, using proprietary NLP for unscripted, dynamic NPC conversations that adapt in real-time.
    – This strategy aims at long-term player retention and infrastructure independence, not just reducing production costs.
    – The scale of this investment sets a new baseline for backend spending in the live-service gaming market.

    Global mobile gaming spending shift begins

    The global mobile gaming industry is undergoing a major shift in spending as Pocket Gamer Connects Barcelona get’s under way. Shanghai publisher HoYoverse plans to invest 14.6 billion dollars into internal generative artificial intelligence infrastructure over the next three years, according to official reports via the media.

    Vertical integration strategy outlined by co-founder

    The strategy, outlined by HoYoverse co-founder Liu Wei during a private presentation in Beijing, focuses entirely on vertical integration. Instead of licensing third-party large language models from commercial tech providers, the studio is establishing a self-contained computing ecosystem. This require building proprietary graphics processing unit clusters and developing custom model-training frameworks for active live-service games.

    First public test for new pipeline

    The publisher will use this framework to automate coding pipelines and deliver real-time content personalization. The first public test for this pipeline is Petit Planet, the company’s upcoming cozy life simulation title that recently concluded its Stardrift closed beta phase on mobile platforms and PC. Traditional life simulation games rely on hardcoded dialogue trees that loop once exhausted.

    Dynamic NPC conversations planned

    HoYoverse plans to deploy proprietary natural language processing models directly into the game logic to allow non-player characters to engage in unscripted, dynamic conversations. These digital villagers analyze player choices and environmental changes on each planetoid, adapting their routines and verbal responses on the fly. This capital allocation seperates HoYoverse from competitors using automated tools purely to reduce production overhead.

    Gamble on player retention and independence

    It represents a direct gamble on long-term player retention and infrastructure independence. Addressing the technical risks inherent to the project, Liu Wei noted to doctoral students that if the pipeline fails to achieve its goals, the studio will view the capital loss as an elaborate fireworks display. As PGC Barcelona begins, the sheer scale of this investment sets a new baseline for live service backend spending.

    Widening gap in platform scaling

    HoYoverse’s massive backend investment underscores a widening gap in how next-generation titles scale across platforms. For a closer look at how these technical demands play out in real-time—including how the latest crossover event pushed Wuthering Waves near 50,000 concurrent Steam players while pushing mobile hardware to its limits.

    • Gameworldobserver.com
    • Hoyoverse.com/Petite Planet
    • Pgconnects.com
    Sources
    Tags: , , , ,
  • Hideo Kojima: AI Won’t Create Real Art in My Lifetime

    Hideo Kojima: AI Won’t Create Real Art in My Lifetime

    Key Takeaway

    – Kojima views generative AI as a tool for speeding up development workflows, not for creating art.
    – He does not believe AI will create “true art” within his lifetime.
    – Kojima emphasizes that humans must remain central to the artistic process; AI is only a “janitor for creative chores.”
    – He is more interested in using AI for control systems than for generating visuals or artwork.
    – It is unlikely his upcoming games (e.g., OD, Physint) will incorporate generative AI-generated artwork or designs.

    Hideo Kojima Clarifies His Stance on AI in New Interview

    Hideo Kojima has recently expressed enthusiam for generative AI, thou his latest comments reveal a more nuaced perspective. In an interview with the Washington Post, he clarified that while AI can speed up development workflows, he dosen’t believe it will play a role in creating ‘true art’ within his lifetime. He discussed this at a high-profile Prada art event held at New York’s Chelsea Hotel, shortly after a promotional short film featuring AI-driven marketing drew widespread critisism.

    AI as a “Janitor” for Creative Chores

    Kojima explained that art is life, but in fifty years or a hundred years, maybe AI could create art, but while he lives, he dosent think he’ll see it. He finds no interest in the prospect of AI-generated art. The Washington Post reported that Kojima views AI as essentially “a janitor for creative chores,” emphasizing that humans must “stay in the room where art gets made.” He further suggested society might find a good way to use the technology, but it will be up to the younger generation to figure that out.

    Evolving Views and Future Projects

    His earlier comments about AI were more ambiguous and conveyed far greater enthusiam. However, this recent interview offers a more grounded view. He previously noted AI would be useful for non-visual purposes, saying he’s more intersted in using AI in control systems rather than for creating visuals. Given these evolving views, it now seems unlikely Kojima will incorporate generative AI-generated artwork in upcoming titles like the Xbox-exclusive horror game OD or the PlayStation-exclusive Physint, which is not expected until 2030.

    The Human Element Remains Paramount

    The human element remains paramount for Kojima, with generative AI being nothing more than a utilitarian tool in his view. The promotional video featuring a digital version of him with Nicolas Winding Refn was labeled as “AI slop” by fans. He addressed the broader conversation surrounding art and AI head-on, explaining history and creativity require a human touch. The technology may evolve, but Kojima’s stance is clear: true art requires a human presence.

    Sources
    Tags: , , , ,
  • KH Collection May Use AI Art for PS5, Switch 2 Covers

    KH Collection May Use AI Art for PS5, Switch 2 Covers

    Key Takeaway

    – Kingdom Hearts 4 trailer shown, plus older games rereleased in a new bundle
    – Bundle cover art accused of using generative AI (e.g., Donald Duck having wrong finger counts)
    – Background details like clock towers and skyscrapers show AI-like inconsistencies
    – Fans believe Tetsuya Nomura didn’t create the AI art; publisher may have recycled assets
    – Critics say Square Enix used AI to save costs, distorting original artwork

    First Impressions of the Kingdom Hearts Collection

    The June Nintendo Direct definitely delivered for fans of the popular Square Enix and Disney RPG. Not only was a new Kingdom Hearts 4 trailer shown but older games is being reintroduced in a new bundle. However the Kingdom Hearts Collection [I~III] is facing scrutiny for allegedly using generative AI artwork.

    Pre-Order Concerns About The Covers

    With an October 8th release date the collection is available for the Switch 2, PS5, and Xbox consoles. Buyers considering pre-ordering the title recognised suspicious-looking retail covers. One of the most obvious examples is with the Switch 2 box where Donald Duck has a different number of fingers on each hand. The background graphics ring even more alarm bells.

    Details on the Suspicious Artwork

    Aspects of the clock tower and skyscraper may be products of genAI tools. Even on the Kingdom Hearts Collection PS5 cover followers detected inconsistencies. The various renditions of Sora depict the series main hero with oddly shaped hands and thumbs. Legendary Square Enix designer Tetsuya Nomura has created the bulk of the artwork for the Kingdom Hearts franchise.

    Fan Theories and Explanations

    Most fans doubt that Nomura himself would resort to machine learning leading them to consider other theories. DekuDraws who is an avid follower of Nomura offered an explanation on social media. He believes the work is authentic noting that on the Switch 2 box one of Donald Ducks mitts is hidden. It’s possible that the publisher used AI tools to separate the illustrations for the alternative covers.

    Possible Shortcuts and Criticisms

    Square Enix may have taken shortcuts to piece together existing illustrations into something new. That would explain the inconsistencies with how the characters positioning. They also don’t seem to blend into the background scenery which likely wasn’t originally drawn with them. At best the publisher is using sloppily designed images to promote the Kingdom Hearts Collection.

    Final Thoughts on The Controversy

    Critics point out that to save costs relying on generative AI to recycle assets can distort an artists work. Tracker_TD@Bluesky and DekuDraws X account has added their voices to the discussion but Square Enix has yet to issue a statement. The debate continues among fans and artists about the ethics of such techniques. This could set a precedent for future video game packaging and promotion. Many hope for clarity before the October release date.


    Sources

    Tags: , , , ,
  • Generative AI in Stellar Blade Trailer May Alienate Chinese Players

    Generative AI in Stellar Blade Trailer May Alienate Chinese Players

    Key Takeaway

    – Garbled AI-generated text in *Stellar Blade: Blood Rain* key art and trailer suggests genAI tool use has drawn criticism.
    – Analyst Daniel Camilo warns that garbled Chinese characters could hurt sales in China, which accounted for over 60% of the original game’s PC audience.
    – Shift Up CEO Kim Hyung-tae supports genAI as a productivity tool for smaller Korean studios, but its final role in the sequel remains unclear.
    – Beyond AI concerns, the sequel faces backlash over its youthful protagonist, Evie, adding further controversy.

    Generative AI has become an increasingly common resource in game development. Regardless, that hasn’t stopped the scrutiny of studios that rely on the technology. Most recently, there have been concerns about the Stellar Blade: Blood Rain trailer that premiered at Summer Game Fest 2026.

    Suspicious Details in the Trailer and Key Art

    The video has several telltale signs that genAI tools influenced its production. Even in the key art used by Shift Up to promote the Stellar Blade sequel, windows on buildings resemble AI slop. Daniel Camilo, on social media, focused on Chinese characters that don’t translate to anything coherent.

    Persistent Flaw of Garbled Text

    While machine learning now produces more convincing images and videos, garbled text remains an issue. It’s possible that the suspicious artwork is a placeholder ahead of the Stellar Blade: Blood Rain release date. Nevertheless, Camilo thinks that it could hurt sales from one of the largest customer bases.

    Formerly a PS5 exclusive, Stellar Blade debuted on Steam in June 2025. The PC port’s success surprised many analysts, with one Asian country significantly boosting sales. Gamalytic revealed that over 60% of the audience of the stylish hack-and-slash RPG hailed from China.

    Multi-Platform Launch and Familiar Setting

    The follow-up will likely launch on multiple platforms, with many of the same buyers anxious for the sequel. Its setting could also attract players who find the scenery familiar. Some fans believe that the city of Chongqing inspired the Cyberpunk-themed trailer.

    Shift Up CEO’s View on Generative AI

    How much of a role generative AI will have in the final build is unclear. However, Shift Up CEO Kim Hyung-tae is a strong supporter of the technology. He previously explained that smaller Korean companies need the tools to compete with larger western and Chinese studios. Still, the executive says it should increase productivity rather than create assets.

    Even if the Chinese characters are overlooked, Blood Rain will not escape controversy. Other critics argue that its youthful protagonist, Evie, is a questionable choice for the sequel.

    Sources
    Tags: , , , ,
  • Ubisoft Tests Generative AI in Far Cry 7

    Ubisoft Tests Generative AI in Far Cry 7

    Key Takeaway

    – Ubisoft is heavily investing in generative AI despite a record financial loss and job cuts, using Far Cry 7 as a research testbed for AI tools.
    – Projects like “Teammates” aim to create memory-enabled, real-time conversational NPCs via Google Gemini, with La Forge leading the effort.
    – The AI push follows past controversial experiments (e.g., NFT Quartz) and hasn’t yet shown clear evidence of impacting final releases; investors remain skeptical.
    – The company reported a €1.3B operating loss for 2025-26, cash support from Tencent, and expects a low free-cash-flow year ahead before any rebound.

    Ubisoft is pushing ahead with aggressive investments in generative AI, even as the gaming giant faces its worst financial year in history. According to a recent report, the company is using an early build of the unannounced Far Cry 7 as a testing ground for these new AI tools.

    Industry Developments

    As first reported by The Gamer, Tom Henderson of Insider Gaming briefly posted on X that the generative AI stuff currently “looks like sh*t” before deleting the post. However, Henderson explained that Far Cry 7 is being used purely for research and development purposes, so fans shouldn’t necessarily expect AI generated content to make it into the game’s final release.

    Financial Snapshot

    This news follows Ubisoft’s recent earnings report, which showed a record operating loss of €1.3 billion in the 2025-26 fiscal year. It’s been a tough ride for the publisher. Net bookings have dropped by over 17 percent down to €1.53 billion, seven game projects were outright canceled, six more were delayed, and roughly 1,200 employees lost their jobs over the past year.

    Strategic Support

    A major €1.16 billion cash injection from a Tencent transaction helped keep the company’s balance sheet somewhat stable, but Ubisoft warned investors that the next fiscal year will be a low point for free cash flow before any projected rebound.

    AI Commitments

    Despite the financial bleeding, Ubisoft is heavily doubling down on artificial intelligence. The publisher said in its earnings report that it is increasing investments in generative AI to build dynamic NPCs and quality assurance bots.

    • One of the big ones is a project called “Teammates,” an R&D experiment built on Google Gemini that seeks to make NPCs into conversational companions that can remember player interactions and respond in real-time. Ubisoft’s La Forge division is heading up the project, and the company has already shown off similar AI tech in partnership with Nvidia and Inworld AI at the Game Developers Conference.

    Historical Context

    For longtime observers of the company, this aggressive push into a controversial new tech trend might feel a bit familiar. Back in late 2021, Ubisoft chased the NFT craze by launching the Quartz platform in Ghost Recon Breakpoint, a move that was ultimately abandoned following intense backlash from players.

    Market Reactions

    Now, they are facing similar skepticism from both gamers and the stock market. When Ubisoft first announced its accelerated investments in AI back in January, the news triggered a massive 34 percent collapse in its share price in a single day, dropping the company’s market cap below €1 billion.

    Whether their expensive bet on AI will actually pay off in future titles or become another abandoned experiment remains to be seen, but for now, the publisher is weathering a historic financial storm while banking heavily on the tech.

    Sources
    Tags: , , ,
  • Battlefield 6 Developers Utilize Generative AI for Game Design

    Battlefield 6 Developers Utilize Generative AI for Game Design

    Key Takeaways

    1. Generative AI has significantly improved voice-over lip-syncing and character face creation in Battlefield 6.
    2. EA faced criticism for questionable DLC releases, including a cosmetics pack with poorly designed items.
    3. AI tools like Voice2Face and FaceRig contributed to the animation and character design processes, with 30% of animated speech generated by AI.
    4. Disclosure on AI usage in games on Steam is not strictly required, leading to potential concerns from players regarding transparency.
    5. EA previously stated that no AI-generated assets would be included in Battlefield 6, raising skepticism among players about their commitment to this claim.

    There is proof that generative AI has been a significant factor in crafting Battlefield 6. The business website Fast Company recently spotlighted the most creative game developers in their article. The report mentioned that voice-over lip-syncing and character face creation saw improvements thanks to this technology. However, it’s debatable if this application requires a clear disclosure on Steam.

    Scrutiny Over DLC

    EA’s shooting game has already been under fire for releasing questionable DLC. One instance included a cosmetics pack with a strangely designed weapon. Moreover, a skin featuring a creepy mask bore a strong resemblance to older artwork from Call of Duty. While some of these items were removed by developers, there was no explanation given regarding their source.

    AI’s Impact Pre-Release

    The article from Fast Company, which was discovered by Insider Gaming, examines the role of AI leading up to the Battlefield 6 launch. Voice2Face is a tool that analyzes spoken words and produces precise lip-synced animations. The report states that “30% of the final animated speech was generated” by this tool.

    Additionally, FaceRig helped artists to tweak characters’ faces for Battlefield 6, making the task easier. Although a lot of the process involved human effort, the end results appear more refined compared to rushed artwork. Nonetheless, players can notice the influence of machine learning in the final game.

    Disclosure on AI Usage

    A game on Steam isn’t strictly required to disclose how AI influenced its creation. Game studios can utilize this technology for idea generation or enhancing coding productivity. But, Valve mandates a disclosure when there’s “Player-facing AI output.” Whenever developers opt for shortcuts in designing characters, some critics are likely to raise objections.

    Publishers are increasingly bending the rules and skipping disclosures, even for minor uses. The most recent issue arose with Crimson Desert, where Pearl Abyss failed to erase placeholder AI images. One frustrated player asked for a refund on Steam after only 4 hours of gameplay, but was unsuccessful.

    Before the launch of Battlefield 6, EA claimed that no assets generated by machines would be in the retail version. Given the company’s large investment in AI technologies, players had their doubts. The gaming community will keep discussing whether the studio upheld its commitment.

    Source:
    Link

     

    Tags: , ,
  • Pearl Abyss Reveals Generative AI Assets in Crimson Desert on Steam

    Pearl Abyss Reveals Generative AI Assets in Crimson Desert on Steam

    Key Takeaways

    1. Pearl Abyss used generative AI to create temporary 2D visual props during early development of Crimson Desert, which were not replaced by launch.
    2. The studio acknowledged their mistake and is reviewing all assets to ensure no other generative AI results are present in the game.
    3. A disclaimer has been added to the game’s Steam page, noting the use of generative AI in creating some 2D prop assets.
    4. Community reactions include criticism over odd artwork and concerns about AI translation errors in the game’s dialogues.
    5. The situation reflects a growing trend of controversies regarding AI usage in game development, with other games facing similar scrutiny.

    After players noticed strange images in Crimson Desert, they accused Pearl Abyss of cutting corners. The developer has responded on social media, stating that it utilized generative AI. It admitted that it did not remove the placeholder images from earlier versions of the open-world game.

    Clarification from Pearl Abyss

    Pearl Abyss explained that the “2D visual props were created during the early stages of development using experimental AI generative tools.” These temporary assets were crucial for the team to experiment with the “tone and atmosphere” of the massive game world. The artists had planned to replace these images by the game’s launch date, but that clearly failed to happen.

    Acknowledgment of Mistake

    The studio has expressed regret over what it describes as a mistake. It is currently reviewing all assets to confirm that no other results from the debated technology were missed. Gamers can expect forthcoming updates to erase any problematic artwork.

    The Steam page for Crimson Desert now includes a disclaimer, as mandated by Valve. It states, “Generative AI technology is used in a supplementary capacity during the creation of some 2D prop assets.” Like the statement on social media, Pearl Abyss points out that these placeholder images were supposed to be temporary.

    Community Reactions

    Reddit users highlighted several instances of framed artwork showing people with limbs in odd positions. There were also suspicions that AI tools had translated dialogues into German and other languages, leading to numerous errors.

    While this explanation might silence some critics, others observe that these types of apologies are becoming quite frequent. A notable case is Clair Obscur: Expedition 33, which, after being awarded Game of the Year, was disqualified by Indie Game Awards when players found proof of AI usage. Similarly, Sandfall also employed placeholders during its development before removing them from the final retail version.

    Crimson Desert is a significant project, with reports suggesting that around 250 developers were involved. Building Pywel was a monumental task, yet some gamers argue that the studio has enough resources to avoid such shortcuts.

    Source:
    Link

     

    Tags: , ,
  • YouTube Asks Viewers to Spot AI-Generated Content in Ratings

    YouTube Asks Viewers to Spot AI-Generated Content in Ratings

    Key Takeaways

    1. YouTube allows and encourages the use of generative AI in video creation but aims to combat “AI slop” through viewer feedback.
    2. A new rating system is being tested, asking viewers to evaluate the quality of AI-generated content.
    3. Creators can enhance their videos using AI tools, but low-quality uploads risk monetization penalties.
    4. YouTube’s checks for content quality via automated and human evaluations have proven insufficient, with many shorts being flagged as poor quality.
    5. Critics warn that the new system may inadvertently increase “AI slop” and allow the platform to train models that create misleading content.

    YouTube’s relationship with generative AI is quite complex. The platform permits certain videos made by machine learning and even encourages their usage. However, they have also pledged to tackle what they call “AI slop.” A new rating system has been introduced, urging viewers to identify this unwanted content.

    Viewer Engagement with AI Content

    VidIQ is one of the several social media accounts that have spotted pop-ups within the YouTube app. It straightforwardly inquires, “Does this feel like AI slop?” or asks if “low-quality AI” played a role. Responses can range from “Not at all” to “Extremely.” As of now, this new approach seems to be in a limited testing stage.

    Content Creation and AI Tools

    Creators are welcome to utilize generative AI tools to improve their videos without needing to record voice-overs, edit, or design graphics. However, an increasing number of uploads are being made with little human supervision. Even so, they might still be accepted unless they are classified as low-quality. If not, channel operators risk losing their ability to monetize.

    Standards and Oversight

    YouTube checks if a candidate fulfills basic requirements through both automated and human evaluations. However, both methods have shown to be insufficient. A recent study indicated that over 20% of YouTube shorts were poorly made, repetitive, or misleading. This might be the reason the company is incorporating a new feature to the standard like/dislike buttons.

    Potential Issues with Viewer Reliance

    Depending on viewers has its flaws, as some may not be able to recognize sophisticated deepfakes. There is also a subjective factor, where fans of a channel might hesitate to report a video.

    Criticism of the New System

    Some critics argue that the new rating system could actually increase the amount of AI slop instead of reducing it. If these changes are widely adopted, users will generate vast amounts of data. TukiFromKL suggests that the platform might be training its own models to create content that’s trickier to identify. With some outputs more convincing than others, it could learn how to effectively deceive viewers.

    Source:
    Link

     

    Tags: , ,
  • Steam Next Fest Overrun with Low-Effort AI Demos, Users Warn Valve

    Steam Next Fest Overrun with Low-Effort AI Demos, Users Warn Valve

    Key Takeaways

    1. Gamers are unhappy with this year’s Steam Next Fest due to a rise in low-quality demos, often created using generative AI or reused assets.
    2. Many demos have been labeled as “AI slop,” indicating a perception of low-effort and generic content among players.
    3. Over 17% of demos in recent Next Fest events have been tagged with Steam’s generative AI label, showing a growing presence of AI-driven games.
    4. Discussions on Steam highlight the distinction between using AI tools and asset flips, raising concerns about game discoverability and curation.
    5. The debate around AI usage in game development is expected to grow, focusing on the balance between transparency and quality on platforms like Steam.

    This year’s Steam Next Fest, which kicked off today (Feb 24), is facing significant backlash from gamers who are calling attention to what they see as a surplus of low-quality demos relying on generative AI or reused assets. The Steam Next Fest events serve as promotional opportunities where numerous developers offer free demos for players to try within a week. This approach is designed to allow smaller studios to gain visibility and for players to preview games before their release. However, many in the community are noticing not just fresh concepts, but an increase in titles that seem hurried, lack polish, or depend too heavily on AI-generated content.

    Community Concerns

    On Steam’s forums and community threads, players have labeled these demos as “AI slop,” a term that has become popular for describing digital content created with AI that appears low-effort, generic, or devoid of meaning. One user mentioned that although they are “enjoying the Next Fest,” the overwhelming presence of “poor-quality AI slop games” is “incredible.” The user further cautioned that if tools for discovering quality titles do not improve, worthwhile games could easily get overlooked.

    Ongoing Issues

    It’s important to note that this concern is not entirely new. Reports indicate an increasing number of demos marked with Steam’s generative AI label. Valve implemented this disclosure to clarify when AI tools were utilized in various aspects like art, writing, or coding. Some estimates suggest that over 17% of demos in recent Next Fest events have included this tag. Gamers are witnessing that AI-driven games are no longer a niche category.

    Broader Implications

    Critics of the current landscape do not place all the blame solely on AI. Discussions on Steam’s boards are exploring the distinction between using AI tools and the larger issue of asset flips—games created from pre-existing components with minimal creative contribution. Some argue that merely labeling games as “AI used” does not tackle the fundamental problem of discoverability and curation on a platform with an almost endless array of listings.

    Previously, Valve’s AI disclosure policy has also sparked conversations beyond Steam Next Fest. For example, Epic Games’ CEO has openly questioned the merit of marking games for AI usage, suggesting that this practice may soon become outdated as these tools are widely available.

    Future of Game Development

    As we look ahead, it’s likely that these discussions will intensify, as finding a balance between transparency and quality is becoming more challenging on an open platform like Steam, especially with generative tools playing a more crucial role in game development.

    Source:
    Link

     

    Tags: , ,
  • Light-Based Chip Offers 100x Faster Performance than Nvidia A100

    Light-Based Chip Offers 100x Faster Performance than Nvidia A100

    Key Takeaways

    1. LightGen is an all-optical computing chip developed by researchers from Shanghai Jiao Tong University and Tsinghua University to support generative AI demands.

    2. The chip features over two million artificial neurons in a compact design, allowing it to perform complex tasks like high-definition video creation and 3D modeling.

    3. LightGen introduces an “optical latent space” that processes high-dimensional data using light, maintaining full-resolution images and improving throughput significantly.

    4. The chip operates over 100 times faster than a top Nvidia A100 GPU, demonstrating its potential for advanced data processing.

    5. While still reliant on external laser setups, LightGen represents a promising shift toward rapid, energy-efficient intelligent computing for the future of generative AI.

    Researchers from Shanghai Jiao Tong University and Tsinghua University have introduced “LightGen,” an innovative all-optical computing chip that is tailored to meet the growing demands of generative artificial intelligence. The chip, which is explained in the journal Science, marks a major transition from electronic transistors to photonic neurons, potentially addressing the significant energy challenges currently faced by the AI sector.

    Major Advancements in Optical Processing

    Unlike older optical processors that had only a few thousand neurons and were mainly used for simpler operations like image classification, LightGen employs sophisticated 3D packaging to incorporate more than two million artificial neurons into a compact quarter-square-inch device. This extensive capacity enables the chip to perform intricate generative tasks, such as creating high-definition videos and 3D models, which were once only possible with advanced electronic GPUs.

    A New Approach to Data Processing

    One of the key innovations in LightGen’s design is the “optical latent space.” By utilizing ultra-thin metasurfaces and arrays of optical fibers, the chip can compress and process high-dimensional data solely through light. This feature allows it to manage full-resolution images without needing to break them into smaller sections, which keeps essential statistical information intact and significantly boosts throughput. The researchers found that the chip operates more than 100 times faster than a top Nvidia A100 GPU.

    Promising Future for Intelligent Computing

    In laboratory evaluations, LightGen managed to carry out high-resolution semantic image generation and 3D manipulation at a quality level that rivals leading electronic neural networks. Although this technology is currently dependent on external laser setups and unique manufacturing methods, it lays down an encouraging foundation for the future of rapid, sustainable, and intelligent computing.

    LightGen paves the way for progress in generative AI, enhancing speed and efficiency, and offers a new direction for research in high-speed, energy-efficient intelligent computing. — Yitong Chen, the primary author of the study.

    Source:
    Link

     

    Tags: , ,