📌 TOPINDIATOURS Breaking ai: MIT Researchers Unveil “SEAL”: A New Step Towards Sel

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

📌 TOPINDIATOURS Breaking ai: Report Warns That AI Is About to Make Your Boss a Pan

While the world holds its breath to see if the US tech industry can bring about human-level artificial intelligence — and therefore completely rewrite the social contract and economy — experts are warning current iterations of the tech are already remaking the workplace in insidious ways.

This week, the European Trade Union Confederation (ETUC), a group representing nearly 45 million workers across 40 European countries, published a comprehensive report about the disturbing impact of algorithms that increasingly control the workplace.

Titled “Negotiating the Algorithm,” the 70-page tome lays out the basic facts about algorithmic management, the use of AI programs to oversee workers on the job. Far from a dystopian fantasy, the ETUC alerts readers to the alarming rise of the tech. In fact, an early study from this year found that 79 percent of job sites across the EU — and 90 percent in the US — already use at least one algorithmic management tool to govern the rank and file.

Though most of us roll our eyes when our bosses roll out a new piece of software at work, algorithmic management is already drastically changing the power dynamics of the workplace. And as the ETUC warns, those changes never seem to pan out in the worker’s favor.

Ultimately, the ETUC guide identifies seven risks — or functions, depending on your point of view — that come from AI governance: discriminatory work assignments, fluctuating wages, loss of worker control, constant surveillance, unreasonable performance evaluations, automated punishment, and non-payment.

“Algorithmic management is used to determine work allocation and pay in ways that are typically opaque and often discriminatory,” the report reads. “Workers have to contend with intensive forms of surveillance which reduce autonomy and undermine privacy. Workers are evaluated in ways that are not transparent and with no opportunity for worker input.”

“Perhaps worst of all,” it continues, “workers face algorithmically-determined punishments, up to and including the loss of their job, sometimes without ever being able to communicate with a human boss.”

While algorithmic management is now built into precarious low-wage jobs such as ride hailing, warehouse labor, and cloudwork, it’s also being rapidly expanded to sectors like therapy, legal work, and healthcare. In other words, just because you’re not making a living off an app now, doesn’t mean you won’t be sometime in the future.

Fortunately, the ETUC offers solutions for fighting the AI panopticon. The first is a handy list of victories workers are already winning across the EU in places like Denmark, where the rent-a-maid app Hilfr agreed to give workers a comprehensive explanation for all algorithmically-determined decisions, building on previously negotiated dignities like a minimum wage and paid sick days.

For cases where the company isn’t so willing to share data about its systems, the report highlights tactics workers can use to crack the algorithm. These start with a polite request for company data according to EU data laws.

If that doesn’t work, there are less official methods to fall back on. These include the sock-puppet technique, where a gig worker creates multiple accounts to compare data against their main profile, reverse engineering, a more labor-intensive peak into the raw data, and “counter apps” like UberCheats, which are used to audit algorithmic management software.

Ultimately, the report concludes, fighting back against algorithmic management isn’t about “reinventing the wheel, it’s about adding on new spokes.”

More on labor: Uber Drivers Say They’re Getting Locked Out of the App and Trapped in a Kafkaesque Limbo When They Try to Dispute It

The post Report Warns That AI Is About to Make Your Boss a Panopticon Overlord appeared first on Futurism.

🔗 Sumber: futurism.com