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

📌 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

📌 MAROKO133 Update ai: Watch: $80,000 humanoid robot takes a beating in wild YouTu

Cody Detwiler, popularly known through his YouTube channel ‘WhistlinDiesel’, tested a humanoid’s endurance capabilities with intense tasks, ultimately destroying the robot. The humanoid in question was a Unitree G1 named Ben, worth $80,000.

The YouTuber released his 17-minute video titled ‘What Happens if you Abuse a Robot’ on the platform on September 30. It has garnered over 2 million views at the time of writing this report.

From smashing into glass doors as a burglar to attacking with a machete and acting as a ‘dangerous’ nanny, the Unitree G1 was initially tested for its usefulness as an assistant before appearing for the durability tests.

Ben was programmed to ‘hate’ humans and consider them enemies in this video. The actions it took as the video culminated showed the hazardous effects of what the machine could probably do if it falls in the wrong hands.

The series of tests

YouTuber Cody Detwiler put Ben the humanoid through a series of tasks after insulting him and programming him to consider humans as enemies.

In one of the tasks, the robot was handed a machete and asked to attack after he was cornered by Detwiler’s team. And, Ben carried out his instructions with precision and without mercy.

In another activity, the humanoid was taught to cook food, albeit for hardly a minute or two. However, it was in a violent mood, and all the food was spilled on the floor while working on it.

Probably the most dangerous part came up when Ben acted as a nanny for baby doll, tasked with taking care of a baby as Detwiler acted to head out for a party.

The robot watched the baby for some time before it turned violent and tried to shove a machete through the crib. Later, it also picked up the baby doll by the leg and started waving it.

The humanoid also acted in a staged burglary and ran at 2m/s up and down on an ascending terrain. Outdoors, WhistlinDiesel drove a truck into Ben, took it into downtown Nashville, and even used it to mow a lawn.

In the end, Ben, the humanoid, was hit by a truck, and its parts were put up for sale.

A concerning future

While the video looked like it was made solely for entertainment purposes, the behavior of the Unitree G1 does raise concerns. Especially in an instance at the start of the video, where it was pushed back rudely.

The humanoid retaliated with aggressive steps in a threatening manner towards Cody Detwiler, almost as if it felt insulted due to his rude behavior.

The manhandling of the baby doll was another concerning situation, although it’s still unknown if the robot recognized it as a doll before taking it out of the crib.

Robots showing emotions is a possibility for now, but looking at the Unitree G1 humanoid, it won’t be long before we live in a world with mechanized counterparts that can feel insults like we do.

🔗 Sumber: interestingengineering.com