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

📌 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: Photo Shows Elon Musk at Jeffrey Epstein Dinner Terbaru 202

A newly revealed photo shows Elon Musk attending a “wild” dinner with Jeffrey Epstein and other powerful tech figures, including Mark Zuckerberg.

The picture, which appears to have been taken by Epstein himself, was contained in the latest tranche of documents released by the Department of Justice, and adds to the growing pile of evidence showing that Musk had much deeper ties to Epstein than he’s let on publicly.

Based on the photo, other files, and previous reporting from Vanity Fair, the dinner — which was already public knowledge — took place on August 2, 2015, and was hosted by LinkedIn co-founder Reid Hoffman.

Epstein mentioned the dinner in an email sent to Peter Attia, a wellness and “longevity” influencer, on that same day, adding that along with Musk, Peter Thiel and Mark Zuckerberg would be in attendance. The next day, he emailed himself a photo taken at the dinner, with Musk and Zuckerberg in clear view.

He evidently thought highly of the occasion. Later that month, Epstein bragged about the dinner in an email sent to Tom Pritzker of the billionaire Pritzker family, calling it “wild.”

An email sent by Epstein’s personal assistant Lesley Groff on the morning of the dinner, with the subject “Reconfirming tomorrow’s dinner,” provides a list of attendees. It includes Musk, Zuckerberg and his wife Priscilla Chan, Hoffman and his wife Michelle Yee, Palantir cofounder Thiel, then-head of MIT Media Lab Joi Ito, MIT neuroscientist Ed Boyden, and Musk’s college roommate Navaid Farooq. 

The dinner was hosted in honor of Boyden by Hoffman, according to Vanity Fair, which first reported on the dinner in 2019. Musk, Hoffman, and Thiel were all considered to be part of the so-called “PayPal mafia,” a group of former heavies at the payment platform that went on to be huge figures in Silicon Valley. 

On top of clearly contradicting Musk’s long-held insistence that he had little contact with Epstein, the dinner revelations also come as the SpaceX CEO has spent the past week or so blasting Hoffman for his apparent ties to the sex trafficker. Shortly after the newly released emails showed that Musk extensively corresponded with Epstein, Musk made a series of tweets highlighting emails claiming Hoffman had visited Epstein’s island and demanded an investigation. Hoffman fired back, sharing evidence of Musk’s own involvement, including an email in which Musk asks Epstein when his “wildest” party will be on the island. Musk continued to maintain his innocence throughout their catfight.

“The big difference between you and me, Reid, is that you went and I did not,” Musk fumed. He added that he “obviously didn’t anticipate anything actually shady, as I was bringing my wife [Talulah Riley] at the time.” 

Notwithstanding that all these emails are taking place well after Epstein was first convicted of sex crimes against underage girls in 2008, some exchanges show that Musk was almost certainly aware, to some degree, of the kinds of parties Epstein was throwing. In a conversation on Christmas day 2012, Epstein cautioned Musk that the “ratio on my island might make Talilah [sic] uncomfortable,” using a euphemism for the number of women, or perhaps young girls, that would be there compared to the number of men. Musk responded: “Ratio is not a problem for Talulah.”

More on Elon Musk: Internet Sleuths Just Found Something About Elon Musk Getting a Massage in the Epstein Files

The post Photo Shows Elon Musk at Jeffrey Epstein Dinner appeared first on Futurism.

🔗 Sumber: futurism.com