📌 TOPINDIATOURS Breaking ai: Union Representing NASA Workers Says Space Agency’s N

During the government shutdown in November, the Trump administration suddenly accelerated its plans to shutter over a dozen buildings and around 100 laboratories at NASA’s iconic Goddard Space Flight Center (GSFC).

The decision angered Democratic lawmakers, who accused them of going ahead with the “consolidation” effort without even consulting them.

In a November 10 letter addressed to Sean Duffy, who served as NASA’s interim administrator at the time, representative Zoe Lofgren (D-CA) noted that her staff had received “disturbing reports that NASA is directing the imminent closure of laboratories and facilities hosting mission-critical capabilities” at the GSFC.

Of particular concern is the campus’ main library, which was unceremoniously shut down last month, as the New York Times reported on December 31. NASA officials tried to downplay these concerns seemingly to no avail — as former and current staffers, advisors, and union representatives continue to watch in horror as the GSFC closures go on.

NASA insiders cried foul following the library’s closure, warning that critical and still-undigitized materials could be thrown out in what they said were reckless efforts by the Trump administration.

NASA’s new administrator, Jared Isaacman, who was confirmed by the Senate on December 17, was angered by the NYT‘s framing, accusing the newspaper of not fully reflecting the “context NASA shared.” He argued that “at no point is NASA ‘tossing out’ important scientific or historical materials.”

That’s despite later admitting that “some materials with no historical or technical value may not be retained” following a “deliberate review” over a period of 60 days.

NASA press secretary Bethany Stevens chimed in as well, describing the moves at Goddard as a “consolidation, not a closure.”

Isaacman’s comments have seemingly done little to reassure rattled NASA staffers. In a January 7 response spotted by Astronomy, Matt Biggs, the president of the International Federation of Professional and Technical Engineers (IFPTE) — a union that represents thousands of NASA scientists and engineers — accused Isaacman of making “patently false” statements.

“The rapid and haphazard shutdown of the library at NASA’s Goddard Space Flight Center, reported on by The New York Times, decimated this valuable collection housed at NASA’s largest research library,” Biggs wrote.

The labor union boss also took issue with Isaacman’s argument that the consolidation efforts were part of the 2022 Goddard Master Plan, which dates back to the Biden administration.

“This was not part of some ‘long-planned facilities consolidation’ as Isaacman claims,” Biggs wrote. “The Goddard Master Plan, written in 2022, does not call for the library’s closure. Building 21, which houses the library, was scheduled for renovation, not elimination.”

Futurism has reached out to NASA for clarification. The agency failed to reply to a previous request for comment regarding details of the 2022 Master Plan.

Biggs also accused Isaacman of misleadingly stating on X that “NASA researchers will continue to have access to the scientific information and resources they need to do their work.”

“That’s simply not true,” he wrote. “Much of the material that was available in the library in Greenbelt, Maryland, is copyrighted or unique out-of-print material that cannot or has not been digitized and will no longer be available to researchers.”

It’s a precarious moment, as NASA’s future continues to be debated in Congress. If it were up to the White House, the historic agency would face the largest budget cuts in its 67-year history. Lawmakers, however, have since passed a counteroffer that would leave the agency’s science budget largely unaffected for the fiscal year of 2026 — which technically started in October, as the chaos has dragged on.

It’s a major test of Isaacman’s leadership after taking control of an agency in crisis last month. Insiders remain highly skeptical of the Trump administration’s approach. They say NASA should be opening doors, not closing them, especially as the United States looks to return astronauts to the Moon in the coming years.

“Where is the consolidation?” Biggs wrote. “The material is not being consolidated with other holdings; it is simply being lost to Goddard and to the broader research community, much of it is being sent to storage or to the dumpster.”

“NASA’s scientists and engineers shouldn’t have to be dumpster divers to do their work,” he added. “We expect better from NASA and its managers.”

More on the closure: NASA Veterans Disgusted by Plans to Shut Down Its Largest Library

The post Union Representing NASA Workers Says Space Agency’s New Administrator Is a Straight-Up Liar appeared first on Futurism.

🔗 Sumber: futurism.com

📌 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