MAROKO133 Hot ai: MIT Researchers Unveil “SEAL”: A New Step Towards Self-Improving AI Terb

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

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

📌 MAROKO133 Hot ai: NASA Shuts Down Voyager 1 Instrument as Its Life Force Fades T

NASA’s Voyager 1 has survived nearly 50 years and a light-day of travel from Earth — but in the end, there’s no question that a deadly combination of entropy and 1970s technology will eventually spell its doom.

With the intrepid little space probe running low on power, engineers at NASA’s Jet Propulsion Laboratory made the decision on Friday to shut down one of its key sensors, a sacrifice that they hope will ensure that the rest of the craft will keep on ticking for at least another year. The sacrifice was the Low-energy Charge Particles experiment (LECP), which has been operating ever since Voyager 1 launched in 1977.

“While shutting down a science instrument is not anybody’s preference, it is the best option available,” said Kareem Badaruddin, Voyager mission manager at JPL, in a NASA announcement.

Voyager 1, like its twin mission Voyager 2, runs on a tiny nuclear battery called a radioisotope thermoelectric generator, providing a gradual drip of warmth that’s converted into electricity. According to NASA, the generators on both probes are now losing power at a rate of four watts per year.

On February 27, however, Voyager 1’s power levels unexpectedly plunged during a routine roll maneuver. Engineers feared that if they dipped any lower, it would trigger an automatic shutdown measure to protect the probe against an electrical fault. Recovering the probe, which is more than 15 billion miles away, after going into this shutdown state would be risky and difficult.

That meant cutting down its power draw. This was a scenario Voyager scientists planned for years ago by agreeing on the order on what parts of the spacecraft they would shut down, and next on the chopping block was the LECP. For nearly 49 years, the instrument has measured low-energy particles that originated from outside the solar system and even beyond our galaxy, illuminating the makeup of interstellar space. 

The Voyager team sent a command to turn off the LECP, but not entirely: a small motor that keeps the instrument spinning will stay online, on the off chance that there’s an opportunity to bring the full thing back online again. It’s not entirely implausible: last year, engineers revived a set of Voyager 1’s thrusters after they had been deemed inoperable for nearly two decades. Of Voyager 1’s original set of ten instruments, two remain online: one that listens to plasma waves, and another that measures magnetic fields. The last instrument to be sacrificed was the cosmic ray subsystem experiment, which engineers shut off in February 2025.

“They are still working great, sending back data from a region of space no other human-made craft has ever explored,” said Badaruddin. “The team remains focused on keeping both Voyagers going for as long as possible.”

The time that shutting down the LECP will buy will allow engineers to prepare a more daring power-saving plan, nicknamed “the Big Bang,” which will involve swapping out a group of powered devices all at the same time, “turning some things off and replacing them with lower-power alternatives to keep the spacecraft warm enough to continue gathering science data.”

The plan will be attempted on Voyager 2 first, which has more spare power, before they try it on Voyager 1. As it stands, the tests are planned for this May and June.

More on space: NASA’s Mars Rover Comes Across Formation That Looks Like the Scales of a Massive Cosmic Reptile

The post NASA Shuts Down Voyager 1 Instrument as Its Life Force Fades appeared first on Futurism.

🔗 Sumber: futurism.com