MAROKO133 Update ai: China to sink servers off Shanghai in world’s first commercial underw

📌 MAROKO133 Breaking ai: China to sink servers off Shanghai in world’s first comme

A Chinese company is preparing to submerge a capsule of servers in the sea off Shanghai in mid-October. The project aims to curb the huge energy costs of traditional data centers and marks one of the world’s first commercial services of its kind.

The world’s websites and apps rely on physical servers. Artificial intelligence has accelerated demand, creating an urgent need for more efficient infrastructure. Data centers on land use energy-heavy cooling systems. By contrast, ocean currents can naturally regulate the temperature of submerged servers.

“Underwater operations have inherent advantages,” said Yang Ye of maritime equipment firm Highlander, which is developing the Shanghai pod with state-owned builders.

Workers have finished constructing the large yellow capsule on a wharf near Shanghai. Once submerged, it will serve clients including China Telecom and a state-owned AI computing company.

Microsoft trialed a similar idea off Scotland in 2018 but never went commercial. In China, the project is part of a government push to cut the carbon footprint of data facilities.

“Underwater facilities can save around 90 per cent of energy consumption for cooling,” said Yang, Highlander’s vice-president.

Powered by renewable energy

Government subsidies are fueling these ventures. Highlander received 40 million yuan (US$5.62 million) for a 2022 project in Hainan, which is still operating.

The Shanghai pod was built in components onshore before being readied for installation at sea. It will draw most of its power from nearby offshore wind farms. Highlander says over 95 per cent of its energy will come from renewable sources.

Zhou Jun, an engineer on the Shanghai project, admitted the build posed challenges. “The actual completion of the underwater data center involved greater construction challenges than initially expected,” he said.

Protecting servers from seawater is critical. Highlander uses a steel capsule coated with glass flakes to resist corrosion. An elevator will connect the underwater pod to a section above water, allowing crews to perform maintenance.

Environmental and technical risks

Experts warn of environmental risks. Heat released by underwater data centers could disturb marine ecosystems.

“The heat emitted could in some cases attract certain species while driving away others,” said Andrew Want, a marine ecologist at the University of Hull, as quoted by SCMP. “These are unknowns at this point – there’s not sufficient research being conducted yet.”

Highlander cited a 2020 assessment of its test project near Zhuhai, which found surrounding water temperatures stayed well below thresholds. But experts say scaling up operations could magnify thermal effects.

“For megawatt-scale data centers underwater, the thermal pollution problem needs to be studied more carefully,” said Shaolei Ren from the University of California, Riverside.

Ren also noted practical hurdles. Laying internet connections between offshore servers and land is more complex than for traditional centers. Researchers in the US and Japan have even warned of vulnerabilities to sound-wave attacks under water.

Despite the hurdles, Ren believes subsea facilities may find a role. “They’re probably not going to replace existing traditional data centers, but can provide service to some niche segments,” he said.

🔗 Sumber: interestingengineering.com

📌 MAROKO133 Update ai: Adobe Research Unlocking Long-Term Memory in Video World Mo

Video world models, which predict future frames conditioned on actions, hold immense promise for artificial intelligence, enabling agents to plan and reason in dynamic environments. Recent advancements, particularly with video diffusion models, have shown impressive capabilities in generating realistic future sequences. However, a significant bottleneck remains: maintaining long-term memory. Current models struggle to remember events and states from far in the past due to the high computational cost associated with processing extended sequences using traditional attention layers. This limits their ability to perform complex tasks requiring sustained understanding of a scene.

A new paper, “Long-Context State-Space Video World Models” by researchers from Stanford University, Princeton University, and Adobe Research, proposes an innovative solution to this challenge. They introduce a novel architecture that leverages State-Space Models (SSMs) to extend temporal memory without sacrificing computational efficiency.

The core problem lies in the quadratic computational complexity of attention mechanisms with respect to sequence length. As the video context grows, the resources required for attention layers explode, making long-term memory impractical for real-world applications. This means that after a certain number of frames, the model effectively “forgets” earlier events, hindering its performance on tasks that demand long-range coherence or reasoning over extended periods.

The authors’ key insight is to leverage the inherent strengths of State-Space Models (SSMs) for causal sequence modeling. Unlike previous attempts that retrofitted SSMs for non-causal vision tasks, this work fully exploits their advantages in processing sequences efficiently.

The proposed Long-Context State-Space Video World Model (LSSVWM) incorporates several crucial design choices:

1. Block-wise SSM Scanning Scheme: This is central to their design. Instead of processing the entire video sequence with a single SSM scan, they employ a block-wise scheme. This strategically trades off some spatial consistency (within a block) for significantly extended temporal memory. By breaking down the long sequence into manageable blocks, they can maintain a compressed “state” that carries information across blocks, effectively extending the model’s memory horizon.
2. Dense Local Attention: To compensate for the potential loss of spatial coherence introduced by the block-wise SSM scanning, the model incorporates dense local attention. This ensures that consecutive frames within and across blocks maintain strong relationships, preserving the fine-grained details and consistency necessary for realistic video generation. This dual approach of global (SSM) and local (attention) processing allows them to achieve both long-term memory and local fidelity.

The paper also introduces two key training strategies to further improve long-context performance:

• Diffusion Forcing: This technique encourages the model to generate frames conditioned on a prefix of the input, effectively forcing it to learn to maintain consistency over longer durations. By sometimes not sampling a prefix and keeping all tokens noised, the training becomes equivalent to diffusion forcing, which is highlighted as a special case of long-context training where the prefix length is zero. This pushes the model to generate coherent sequences even from minimal initial context.
• Frame Local Attention: For faster training and sampling, the authors implemented a “frame local attention” mechanism. This utilizes FlexAttention to achieve significant speedups compared to a fully causal mask. By grouping frames into chunks (e.g., chunks of 5 with a frame window size of 10), frames within a chunk maintain bidirectionality while also attending to frames in the previous chunk. This allows for an effective receptive field while optimizing computational load.

The researchers evaluated their LSSVWM on challenging datasets, including Memory Maze and Minecraft, which are specifically designed to test long-term memory capabilities through spatial retrieval and reasoning tasks.

The experiments demonstrate that their approach substantially surpasses baselines in preserving long-range memory. Qualitative results, as shown in supplementary figures (e.g., S1, S2, S3), illustrate that LSSVWM can generate more coherent and accurate sequences over extended periods compared to models relying solely on causal attention or even Mamba2 without frame local attention. For instance, on reasoning tasks for the maze dataset, their model maintains better consistency and accuracy over long horizons. Similarly, for retrieval tasks, LSSVWM shows improved ability to recall and utilize information from distant past frames. Crucially, these improvements are achieved while maintaining practical inference speeds, making the models suitable for interactive applications.

The Paper Long-Context State-Space Video World Models is on arXiv

The post Adobe Research Unlocking Long-Term Memory in Video World Models with State-Space Models first appeared on Synced.

🔗 Sumber: syncedreview.com