MAROKO133 Update ai: World’s smallest AI computer: NVIDIA’s $3,999 DGX Spark packs 1,000-t

📌 MAROKO133 Update ai: World’s smallest AI computer: NVIDIA’s $3,999 DGX Spark pac

NVIDIA’s long-awaited AI developer-focused mini PC is finally ready to hit the market. After months of preorders, the company will begin selling the DGX Spark on Wednesday, October 15, through NVIDIA.com and select retailers.

Although it looks like a compact desktop, the 2.6-pound DGX Spark isn’t designed for everyday consumers. It’s an AI powerhouse built to help developers, research scientists, and students run advanced models locally.

NVIDIA describes it as “the world’s smallest AI supercomputer,” promising data-center-class performance in a form factor that fits on a desk.

At the heart of the DGX Spark is NVIDIA’s new GB10 Grace Blackwell Superchip. The processor combines a 20-core Arm-based Grace CPU with a Blackwell GPU carrying the same CUDA cores as the RTX 5070 graphics card.

NVIDIA has optimized this setup for desktop AI development, allowing users to fine-tune and run large models without relying on remote cloud access.

The GB10 delivers up to 1,000 trillion operations per second of AI compute, thanks to fifth-generation Tensor Cores and FP4 support.

The system also features NVLink-C2C interconnect technology, which offers five times the bandwidth of PCIe Gen 5. This allows seamless data movement between the CPU and GPU, making it ideal for memory-heavy workloads such as model inference, robotics simulation, and generative AI tasks.

NVIDIA has equipped the DGX Spark with 128GB of LPDDR5x memory, shared between the CPU and GPU, and 4TB of NVMe storage. Connectivity options include four USB-C ports, Wi-Fi 7, and an HDMI connector.

The device runs NVIDIA’s DGX OS, a custom version of Ubuntu Linux preloaded with AI software and developer tools.

Designed for portability and performance

The DGX Spark is small enough to fit in a backpack yet powerful enough to run modern AI reasoning models. NVIDIA specifically highlights compatibility with its own foundation models, including Cosmos Reason for world modeling and GR00T N1 for robotics.

Developers can use the system for training, fine-tuning, and deploying these models without the latency or cost of cloud-based computation.

Weighing just 2.6 pounds, the device runs on standard power from any wall outlet. NVIDIA envisions it as a local workstation that can also connect easily to DGX Cloud or other accelerated infrastructure when scaling up workloads.

The company’s full-stack AI platform ensures users can transition projects between desktop and data center environments with minimal code changes.

However, this power comes at a premium. NVIDIA has priced the DGX Spark at $3,999, excluding local taxes or tariffs. The product will be available directly from NVIDIA and select third-party partners.

Larger DGX Station on the way

The DGX Spark isn’t launching alone. NVIDIA is also preparing the DGX Station, a full-sized desktop tower featuring the more powerful GB300 Grace Blackwell Ultra chip.

While pricing hasn’t been revealed, the company plans to release it later this year in partnership with Asus, Boxx, Dell, HP, and Supermicro.

Together, the DGX Spark and DGX Station mark NVIDIA’s growing push to make AI computing more accessible to individual researchers and small labs. While the Spark may not be meant for casual users, it signals the company’s vision of bringing supercomputer-class AI power to the desktop.

🔗 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