MAROKO133 Eksklusif ai: Adobe Research Unlocking Long-Term Memory in Video World Models wi

📌 MAROKO133 Hot ai: Adobe Research Unlocking Long-Term Memory in Video World Model

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

📌 MAROKO133 Breaking ai: 7 of the world’s most powerful tidal turbines generating

Tides are predictable years in advance, unlike wind or sunlight. Because of this reliability, harnessing the energy of ocean tides has long appealed to engineers. The difficulty lies in building machines large enough to capture meaningful amounts of power while surviving one of the harshest environments on Earth. 

Only a handful of tidal turbines have reached megawatt-scale capacity, and even fewer have delivered sustained, grid-connected performance at sea. This list highlights some of the largest tidal turbines developed to date, including both operational systems and landmark prototypes that shaped today’s technology.

1) Orbital O2 – 2 MW Floating Tidal Turbine, Scotland, UK

The Orbital O2 is widely regarded as the most powerful operational tidal turbine currently in service. Commissioned in July 2021, this floating tidal generator is installed at the European Marine Energy Centre (EMEC)’s Fall of Warness test site off the Orkney Islands and is grid-connected via subsea cable to the local electricity network. 

It features a 2 MW nameplate capacity, achieved through two 1 MW rotors mounted on a 242 feet (74 meter) floating superstructure, which is moored in fast-flowing tidal currents. A key innovation is its retractable leg design, which provides surface maintenance access and reduces the need for heavy vessels. 

The 65 feet (20 meter) diameter rotors provide a large swept area for capturing tidal kinetic energy at speeds exceeding 3 m/s, and the device is expected to operate for up to 15 years. Building on earlier prototypes, O2’s success demonstrates that large floating tidal turbines can reliably deliver predictable renewable power into a real grid.

2) ScotRenewables SR2000 – 2 MW Floating Tidal Turbine, Scotland, UK

The SR2000 was a pioneering 2 MW floating tidal turbine developed by ScotRenewables Tidal Power (now Orbital Marine Power) and tested at EMEC’s Fall of Warness site starting in late 2016. As a full-scale prototype, it was engineered to demonstrate utility-class tidal energy generation and successfully operated in harsh North Atlantic conditions. 

Over its testing programme in 2017-2018, the SR2000 achieved full-rated output, exported power to the local grid, and generated in excess of 3 GWh of renewable electricity over approximately 12 months, a level of output that exceeded the cumulative generation previously recorded across Scotland’s wave and tidal sector. It also endured sea states with waves over 13 feet (4 meters) and maintained generation power during winter storms. 

At times, it supplied up to 25 percent of the Orkney Islands’ electricity demand during continuous generation periods. The machine was removed in September 2018 to make way for the next-generation Orbital O2 turbine, marking the SR2000 as a historic milestone in tidal turbine engineering.

3) SIMEC Atlantis AR2000 – 2 MW-Rated Single-Rotor Tidal Turbine (Design Context)

The SIMEC Atlantis AR2000 is a 2 MW-rated tidal turbine design representing one of the largest single-rotor platforms promoted in the tidal energy sector. Developed by SIMEC Atlantis Energy, the AR2000 was unveiled with design specifications targeting 2 MW output, and its scale positions it among the highest-capacity individual tidal turbines proposed. 

While not yet widely deployed as a grid-connected, operational single unit at the time of reporting, SIMEC Atlantis highlighted the AR2000’s large rotor diameter and output potential as a next step in scaling tidal stream energy beyond earlier 1.5 MW designs like the AR1500. This turbine’s rating reflects industry efforts to push the limits of tidal turbine capacity and contributes to broader device scaling in tidal power markets.

4) AR1500 (MeyGen) – 1.5 MW seabed tidal turbine, Scotland, UK

The AR1500 is a 1.5-MW tidal stream turbine deployed at Scotland’s MeyGen project in the Inner Sound of the Pentland Firth, which is one of the most extensively studied tidal stream projects globally. These are 1.5-MW rated turbines with 18-meter rotor diameters, installed on seabed foundations in high-velocity tidal channels. 

The design uses pitch control to maintain output above a rated flow speed and a yaw module to reorient between ebb and flood tides. In practice, AR1500-class machines helped establish multi-turbine, grid-export tidal generation at utility scale, making them a benchmark for modern tidal deployments.

5) Minesto Dragon 12 – 1.2 MW tidal “kite” turbine, Faroe Islands

Minesto’s Dragon 12 is a utility-scale tidal “kite” rated at 1.2 MW and designed to generate energy by flying on a controlled trajectory underwater. Instead of relying on a fixed seabed tower, the system harvests energy by flying a controlled underwater trajectory.

Minesto reports the system was commissioned in February 2024 at the Vestmanna site in the Faroe Islands and delivered its first electricity to the national grid on February 9, 2024. The company describes Dragon 12 as a 12-meter-wide, 28-ton subsea kite tethered to the seabed, operated through an “8-shaped” flight path to increase effective flow speed across its turbine.

6) SeaGen – 1.2 MW pioneering commercial tidal turbine, Northern Ireland, UK

SeaGen was one of the most important early commercial-scale tidal turbines, installed in Strangford Lough and rated at 1.2 MW using two 600-kW turbines on a pile-mounted structure. It was commissioned in 2008 and decommissioned in 2019, with industry reports confirming its 1.2-MW capacity. 

Project documentation notes a total investment of around £12 million, reflecting the cost of proving a large tidal device in a real marine environment. While smaller than today’s <a href="https://interestingengineering.com/lists/2…

Konten dipersingkat otomatis.

🔗 Sumber: interestingengineering.com