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

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

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: Terrified Investors Are Bracing for an AI Bubble “Reckoning” E

Fears over a growing AI bubble that could wipe out the entire economy if it were to burst continue to mount.

For many months now, investors and even tech leaders have openly been discussing the possibility — but if or when such a collapse could take place remains a subject of heated debate.

Nonetheless, investors are already preparing for that type of major tech sell-off, the Financial Times reports. The reporting shows that plenty of fear and uncertainty remain over the untold billions of dollars being poured into wildly unprofitable AI ventures — a dynamic that’s seen AI companies’ valuations skyrocket to record heights, despite dubious prospects of ever turning a profit.

Some are pulling back on their investments in major tech stocks, while others are outright betting on eventual drops in share prices.

“Whether there are excesses… in the equity market on AI is no longer questionable, but to figure out which exact companies will be the losers and when this reckoning will happen is difficult,” fund management firm Amundi chief investment officer Vincent Mortier told the FT.

One investment fund, Blue Whale Growth, sold its Microsoft and Meta stock in the second quarter of last year, with chief investment officer Stephen Yiu telling the newspaper that “we are concerned about the return on investment in some cases, while some of the valuations are insane — especially in private markets.”

GQG Partners chair Rajiv Jain added that “AI’s massive cash burn remains elevated with very little profitability in sight.” His fund sold all of its Magnificent Seven — investor shorthand for Alphabet, Amazon, Apple, Tesla, Meta Platforms, Microsoft, and Nvidia — holdings.

“Though we had less exposure to a few Mag 7 names throughout 2025, we exited our remaining positions by early November because the risks of an AI bubble blow-up are growing, in our view,” he said.

At the same time, many remain unafraid of an imminent collapse.

“We don’t believe that we are in a bubble,” BlackRock international chief investment officer Helen Jewell told the FT, “but investors should prepare for a bumpy ride in 2026.”

Despite plenty of concerns among investors, banks remain optimistic about future growth. Wall Street has predicted double-digit gains this year. The S&P 500 has surged a whopping 92 percent since October 2022, posting double-digit returns for three years in a row now.

“2026 should be another strong year for AI stocks, with capex likely to surpass expectations,” JPMorgan’s Dubravko Lakos-Bujas wrote in a memo.

Others are far more muted about the outlook. Three years on from OpenAI’s launch of ChatGPT, some are starting to wonder how long the enormous hype surrounding AI can continue to be sustained.

In a 2025 retrospective posted on Monday, Bridgewater hedge fund founder Ray Dalio warned that the tech market is “now in the early stages of a bubble.”

But a crash, some argue, may still be some ways away thanks to plenty of remaining excitement.

“A bubble likely crashes on a bear market,” Cetera Financial Group chief investment officer Gene Goldman told Bloomberg. “We just don’t see a bear market anytime soon.”

More on the AI bubble: AI Investors Furious at Suggestion That There’s an AI Bubble

The post Terrified Investors Are Bracing for an AI Bubble “Reckoning” appeared first on Futurism.

🔗 Sumber: futurism.com