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

📌 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

📌 MAROKO133 Eksklusif ai: You Will Laugh Out Loud When You Hear What the Tech Indu

The AI industry has been pouring untold resources into building out enormous data centers across the world.

The plants are immensely resource-hungry, sucking up huge amounts of fresh water to cool ripping-hot computer hardware. They’re turning into a massive strain on the electric grid, forcing some utility operators to enact rolling blackouts during heat waves and cold weather.

The issue reached a fever pitch after the Washington Post reported that a recent rise in customer energy bills was attributable to the enormous and growing power demands of AI data centers.

In short, it’s no wonder that small towns across the nation are coordinating efforts to keep data centers out — a PR disaster tarnishing a major push by AI companies to scale up their expansive operations.

And it seems like Mark Zuckerberg’s Meta, which has committed to spend $600 billion on AI data centers, is painfully aware of the pushback. As the New York Times reports, the company has already spent $6 million on TV ads to convince Americans that data centers aren’t that bad. As one “folksy” ad showing off a new data center in Altoona, Iowa, argued, “we’re bringing jobs here.”

And it’s not just Meta trying to distract the public from all of the glaring downsides of data centers propping up across the country. Amazon is running its own similar ad campaign in Virginia, for instance, admonishing viewers that the facilities help “connect us to the entire world.”

According to the Financial Times, data center operators are “planning to go on the offensive with a lobbying blitz” as well, trying to get ahead of the growing public backlash. One data center executive told the FT that lobbying spending is a flash in the pan compared to the tens of billions being spent on infrastructure.

“If we’re going to spend tens of billions of dollars this year on capital projects, we probably should spend tens of millions of dollars on messaging,” they argued.

Yet the growing backlash is already hampering construction efforts. Over two dozen projects have already been blocked or delayed this month alone, according to research firm MacroEdge, compared to just 25 total in 2025.

In short, it’s not surprising to see tech giants trying to control the narrative by pouring millions of dollars into changing the public’s opinion about the facilities.

As marketing analysts told the NYT, Meta’s efforts are likely not just aimed at residents. They’re intended to influence lawmakers and policymakers, ensuring that the aggressive nationwide push to build out AI infrastructure won’t become an issue in Washington, DC.

Rising energy bills have already turned out to be a contentious topic, particularly in the greater context of rising costs of living, with senator Chris Van Hollen (D-MA) introducing a bill earlier this month to regulate data center energy use.

Even president Donald Trump, who has been a major proponent of AI, argued that Big Tech must “pay their own way,” in a recent post on Truth Social, suggesting the topic isn’t entirely divided by party lines. Republican lawmakers have also called for more AI regulation and a more careful approach to data center buildouts, as NPR reports.

Other politicians worry about overindexing on AI infrastructure buildouts, which could end badly in the long term if the AI sector were to crash, as many experts have warned it could.

“What I very much worry about with this ad campaign is localities committing to this industry and then saying in ten years, ‘What have we done to ourselves,’” Diane Papan, a Democratic state assemblywoman in California, told NYT.

More on data centers: Trump’s Huge AI Project Is Running Into a Major Financial Problem

The post You Will Laugh Out Loud When You Hear What the Tech Industry Is Spending a Swimming Pool’s Worth of Money to Convince the Public appeared first on Futurism.

🔗 Sumber: futurism.com