📌 TOPINDIATOURS Breaking ai: NVIDIA cleared to sell advanced AI chips to China und

The U.S. is easing restrictions on advanced AI chip exports to China, opening a path for NVIDIA and Advanced Micro Devices to seek approval to sell high-performance processors under revised federal rules.

Under new regulations released Tuesday by the U.S. Commerce Department, chipmakers can apply to ship certain advanced computing processors to Chinese customers on a case-by-case basis.

The move marks a clear departure from Washington’s earlier stance, which presumed rejection of nearly all such requests.

The updated rule places NVIDIA’s H200 and AMD’s MI325X under a new licensing framework overseen by the Commerce Department’s Bureau of Industry and Security.

Instead of blanket denials, applications will now be reviewed individually, provided companies meet a series of strict conditions.

One key requirement is that exporters must demonstrate there is no shortage of the processors in the U.S. and certify that shipments to China will not divert manufacturing capacity away from domestic customers.

The revised policy also applies to exports bound for Macau.

Eligibility is tightly defined by performance thresholds, with only chips operating below specific limits qualifying for case-by-case review. These include processors with a total processing performance below 21,000 and total DRAM bandwidth under 6,500 gigabytes per second.

The Commerce Department made clear that the easing does not apply to military or sensitive uses.

Export approvals will be denied if the chips are intended for military, military-intelligence, nuclear, missile, or chemical and biological weapons applications, or if restricted entities are involved.

Any transaction linked to prohibited end users remains blocked under existing export control rules.

Together, the conditions are intended to prevent advanced U.S. AI hardware from strengthening China’s defense or intelligence capabilities, even as commercial access is partially restored.

Export controls loosen carefully

The revised framework also limits the scale of AI chip shipments to China.

Companies will be allowed to export not more than 50 percent of the total volume of eligible processors shipped to customers in the U.S. market, ensuring domestic supply remains protected.

Exporters must also implement rigorous Know Your Customer procedures to prevent unauthorized use or remote access to the technology.

In addition, all approved chips will be required to undergo independent, third-party testing within the U.S. before they can be shipped.

The regulation follows President Donald Trump’s decision last month to allow U.S. chipmakers to resume limited sales of advanced AI processors to China.

It marks a significant shift from export controls introduced in 2022, which aimed to block Beijing from accessing the most powerful US semiconductor technologies.

China access resumes cautiously

If approved, NVIDIA’s H200 would become the most advanced AI chip legally exported to Chinese customers. Introduced more than two years ago, the processor sits below NVIDIA’s newer Blackwell generation, which remains restricted to the U.S. and allied markets.

NVIDIA is preparing to transition to an even faster chip family named after astronomer Vera Rubin. Those future processors are expected to remain off-limits to China under existing export control rules.

While the revised policy does not guarantee approvals, it lowers the barrier for U.S. companies seeking licenses to sell AI hardware into China’s massive technology market. Each application will be reviewed individually, giving regulators greater flexibility while retaining tight oversight.

For U.S. chipmakers, the change opens a limited commercial channel back into China. For Washington, it reflects a careful recalibration of export controls aimed at balancing economic interests with national security concerns.

🔗 Sumber: interestingengineering.com

📌 TOPINDIATOURS Hot ai: Adobe Research Unlocking Long-Term Memory in Video World M

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