MAROKO133 Update ai: NVIDIA debuts Rubin platform at CES 2026, delivering 50 petaflops, fa

📌 MAROKO133 Update ai: NVIDIA debuts Rubin platform at CES 2026, delivering 50 pet

NVIDIA used the CES 2026 stage today to formally launch its new Rubin computing architecture, positioning it as the company’s most advanced AI hardware platform to date.

CEO Jensen Huang said Rubin has already entered full production and will scale further in the second half of the year, signaling NVIDIA’s confidence in demand.

Huang framed Rubin as a direct response to the explosive growth in AI workloads, particularly large-scale training and long-horizon reasoning tasks. He told the audience that AI computation must continue to rise at an unprecedented pace.

Rubin replaces NVIDIA’s Blackwell architecture, which itself succeeded Hopper and Lovelace.

The update continues NVIDIA’s rapid hardware cadence that has helped turn the company into the world’s most valuable corporation.

Built for agentic AI

Rubin takes its name from astronomer Vera Florence Cooper Rubin and introduces a six-chip architecture designed to work as a unified system.

At the center sits the Rubin GPU, supported by major upgrades to interconnect and storage components.

NVIDIA redesigned NVLink to address communication bottlenecks across large clusters. The company also expanded its BlueField data processing platform to manage growing memory demands from advanced AI systems.

A new Vera CPU joins the architecture and targets agentic reasoning workloads.

NVIDIA designed it to support AI systems that plan, remember context, and act over longer periods.

Rubin systems are already scheduled for deployment across the AI ecosystem. Cloud providers, including partners such as Anthropic, OpenAI, and Amazon Web Services, plan to adopt the platform.

NVIDIA also confirmed Rubin will power HPE’s Blue Lion supercomputer and the upcoming Doudna system at Lawrence Berkeley National Laboratory.

Speed and efficiency gains

NVIDIA claims Rubin delivers major performance improvements over Blackwell. Internal tests show up to 3.5 times faster training performance and 5 times faster inference speeds.

Peak performance reaches 50 petaflops.

Efficiency also improves sharply. Rubin supports up to eight times more inference compute per watt, according to NVIDIA.

These gains matter as AI infrastructure strains power grids and data center budgets worldwide.

Huang has previously estimated that global AI infrastructure spending could reach $3 trillion to $4 trillion over five years. Rubin appears designed for that scale.

Reinventing AI storage

Alongside Rubin, NVIDIA unveiled a new AI-native storage approach powered by BlueField-4.

The Inference Context Memory Storage Platform targets a growing problem in AI systems: managing massive key-value caches.

As AI agents handle long conversations and multi-step reasoning, they generate context data that cannot remain on GPUs indefinitely.

NVIDIA’s new platform extends memory capacity beyond the GPU while maintaining high-speed access.

Senior director Dion Harris said new AI workflows place intense pressure on memory systems.

NVIDIA responded by adding an external storage tier optimized for scaling context efficiently.

The platform uses Spectrum-X Ethernet and NVIDIA’s DOCA framework to share context across rack-scale systems. NVIDIA claims up to five times higher token throughput and power efficiency compared with traditional storage.

BlueField-4-based systems will reach partners in the second half of 2026, as NVIDIA pushes deeper into the full AI infrastructure stack.

We’re at the CES 2026!

You can explore all CES 2026 stories and coverage from the IE team by clicking here.

🔗 Sumber: interestingengineering.com

📌 MAROKO133 Hot ai: MIT Researchers Unveil “SEAL”: A New Step Towards Self-Improvi

The concept of AI self-improvement has been a hot topic in recent research circles, with a flurry of papers emerging and prominent figures like OpenAI CEO Sam Altman weighing in on the future of self-evolving intelligent systems. Now, a new paper from MIT, titled “Self-Adapting Language Models,” introduces SEAL (Self-Adapting LLMs), a novel framework that allows large language models (LLMs) to update their own weights. This development is seen as another significant step towards the realization of truly self-evolving AI.

The research paper, published yesterday, has already ignited considerable discussion, including on Hacker News. SEAL proposes a method where an LLM can generate its own training data through “self-editing” and subsequently update its weights based on new inputs. Crucially, this self-editing process is learned via reinforcement learning, with the reward mechanism tied to the updated model’s downstream performance.

The timing of this paper is particularly notable given the recent surge in interest surrounding AI self-evolution. Earlier this month, several other research efforts garnered attention, including Sakana AI and the University of British Columbia’s “Darwin-Gödel Machine (DGM),” CMU’s “Self-Rewarding Training (SRT),” Shanghai Jiao Tong University’s “MM-UPT” framework for continuous self-improvement in multimodal large models, and the “UI-Genie” self-improvement framework from The Chinese University of Hong Kong in collaboration with vivo.

Adding to the buzz, OpenAI CEO Sam Altman recently shared his vision of a future with self-improving AI and robots in his blog post, “The Gentle Singularity.” He posited that while the initial millions of humanoid robots would need traditional manufacturing, they would then be able to “operate the entire supply chain to build more robots, which can in turn build more chip fabrication facilities, data centers, and so on.” This was quickly followed by a tweet from @VraserX, claiming an OpenAI insider revealed the company was already running recursively self-improving AI internally, a claim that sparked widespread debate about its veracity.

Regardless of the specifics of internal OpenAI developments, the MIT paper on SEAL provides concrete evidence of AI’s progression towards self-evolution.

Understanding SEAL: Self-Adapting Language Models

The core idea behind SEAL is to enable language models to improve themselves when encountering new data by generating their own synthetic data and optimizing their parameters through self-editing. The model’s training objective is to directly generate these self-edits (SEs) using data provided within the model’s context.

The generation of these self-edits is learned through reinforcement learning. The model is rewarded when the generated self-edits, once applied, lead to improved performance on the target task. Therefore, SEAL can be conceptualized as an algorithm with two nested loops: an outer reinforcement learning (RL) loop that optimizes the generation of self-edits, and an inner update loop that uses the generated self-edits to update the model via gradient descent.

This method can be viewed as an instance of meta-learning, where the focus is on how to generate effective self-edits in a meta-learning fashion.

A General Framework

SEAL operates on a single task instance (C,τ), where C is context information relevant to the task, and τ defines the downstream evaluation for assessing the model’s adaptation. For example, in a knowledge integration task, C might be a passage to be integrated into the model’s internal knowledge, and τ a set of questions about that passage.

Given C, the model generates a self-edit SE, which then updates its parameters through supervised fine-tuning: θ′←SFT(θ,SE). Reinforcement learning is used to optimize this self-edit generation: the model performs an action (generates SE), receives a reward r based on LMθ′’s performance on τ, and updates its policy to maximize the expected reward.

The researchers found that traditional online policy methods like GRPO and PPO led to unstable training. They ultimately opted for ReST^EM, a simpler, filtering-based behavioral cloning approach from a DeepMind paper. This method can be viewed as an Expectation-Maximization (EM) process, where the E-step samples candidate outputs from the current model policy, and the M-step reinforces only those samples that yield a positive reward through supervised fine-tuning.

The paper also notes that while the current implementation uses a single model to generate and learn from self-edits, these roles could be separated in a “teacher-student” setup.

Instantiating SEAL in Specific Domains

The MIT team instantiated SEAL in two specific domains: knowledge integration and few-shot learning.

• Knowledge Integration: The goal here is to effectively integrate information from articles into the model’s weights.
• Few-Shot Learning: This involves the model adapting to new tasks with very few examples.

Experimental Results

The experimental results for both few-shot learning and knowledge integration demonstrate the effectiveness of the SEAL framework.

In few-shot learning, using a Llama-3.2-1B-Instruct model, SEAL significantly improved adaptation success rates, achieving 72.5% compared to 20% for models using basic self-edits without RL training, and 0% without adaptation. While still below “Oracle TTT” (an idealized baseline), this indicates substantial progress.

For knowledge integration, using a larger Qwen2.5-7B model to integrate new facts from SQuAD articles, SEAL consistently outperformed baseline methods. Training with synthetically generated data from the base Qwen-2.5-7B model already showed notable improvements, and subsequent reinforcement learning further boosted performance. The accuracy also showed rapid improvement over external RL iterations, often surpassing setups using GPT-4.1 generated data within just two iterations.

Qualitative examples from the paper illustrate how reinforcement learning leads to the generation of more detailed self-edits, resulting in improved performance.

While promising, the researchers also acknowledge some limitations of the SEAL framework, including aspects related to catastrophic forgetting, computational overhead, and context-dependent evaluation. These are discussed in detail in the original paper.

Original Paper: https://arxiv.org/pdf/2506.10943

Project Site: https://jyopari.github.io/posts/seal

Github Repo: https://github.com/Continual-Intelligence/SEAL

The post MIT Researchers Unveil “SEAL”: A New Step Towards Self-Improving AI first appeared on Synced.

🔗 Sumber: syncedreview.com