MAROKO133 Eksklusif ai: Manhattan-sized basalt field in Brazil becomes live testbed for ca

📌 MAROKO133 Eksklusif ai: Manhattan-sized basalt field in Brazil becomes live test

In Brazil, a Manhattan-sized carpet of basalt is turning farmland into a living lab for carbon removal.

Here, rain, soil, and rock are working together to pull carbon from the air—a natural process, accelerated by science.

The project, led by carbon removal company Terradot in collaboration with Microsoft, is testing how Enhanced Rock Weathering (ERW) can remove CO₂ from the atmosphere while benefiting farmers.

For billions of years, silicate rocks have helped regulate Earth’s climate by capturing CO₂ and locking it away in soils and oceans.

Terradot speeds up that geologic process by spreading finely milled basalt on farmland, where it reacts with rain and soil to form bicarbonate, storing carbon in dissolved form.

Brazil offers ideal conditions for ERW: a favorable climate, abundant renewable energy, and accessible basalt from extensive quarry networks.

Over the past year, Terradot has scaled deployment, applying more than 100,000 tonnes of basalt across 4,500 hectares, which is about the size of Manhattan.

The company says its goal is “not only to deliver carbon removal, but also to integrate seamlessly with farm operations and provide tangible agronomic benefits to farmers.”

Microsoft’s support extends beyond a standard offtake deal. The company is backing Terradot’s “measurement-first approach” and funding the field trials, lab work, and data infrastructure that power its verification platform.

“By bringing not just capital but also technical expertise, Microsoft is helping move ERW from promise to practice while ensuring scientific integrity,” Terradot said.

Soil to stream science

At the heart of this work is Sentinel, a soil-to-stream research site built within a commercial-scale deployment in São Paulo state.

“Sentinel shows that ERW can deliver today while generating data that powers the Terradot platform, ensuring rigorous science is embedded into every commercial project,” the company said.

Located on farmland within a single watershed, Sentinel tracks carbon’s full journey: from basalt applied to fields through soils and aquifers to the streams where groundwater resurfaces.

The site features deep groundwater wells, in-situ soil sensors, and surface-water stations to close the loop on how carbon moves through the environment.

With Microsoft’s support, Terradot has drilled soil and sediment cores to bedrock to analyze mineral and chemical profiles.

The research focuses on three areas: how agricultural practices like tillage and fertilizer use affect ERW; how dissolved inorganic carbon travels below the root zone; and what happens when carbon-enriched groundwater resurfaces.

“Together, these efforts move us beyond shallow soil measurements toward a fuller picture of the carbon balance,” Terradot said.

“That deeper perspective allows Sentinel to transform open questions into evidence that strengthens and scales ERW.”

From pilot to proof

Terradot’s next project, Carcará, in Paraná state, has been validated under Isometric’s Enhanced Weathering in Agriculture protocol, with first verified credits expected in Q4 2025.

As the company expands, it is working toward “infrastructure-grade carbon removal”, with plans afoot for large-scale, rigorously monitored projects that connect with existing agricultural and infrastructure networks.

🔗 Sumber: interestingengineering.com

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

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