📌 TOPINDIATOURS Eksklusif ai: US jet-powered Avenger drone achieves autonomous air

A key test for the future of air combat played out in the skies as a General Atomics MQ-20 Avenger autonomous drone intercepted a crewed aggressor aircraft during a live air combat exercise. 

The event, held on January 18,  showed how far combat drone software and sensors have progressed as air forces push toward autonomous systems that can fight alongside human pilots. 

The exercise focused on decision-making, flight control, and airspace discipline, all under realistic conditions spanning miles of airspace and thousands of feet of altitude.

A proving flight for next-generation combat drones

Militaries around the world are investing heavily in combat drones, especially systems designed to operate with limited human input. Building the airframe has proven easier than mastering the software and sensors that allow an unmanned aircraft to find, track, and engage an enemy. Remote control flight has existed for decades, but modern combat requires drones that can process data and react in real time.

General Atomics Aeronautical Systems, Inc. has been using the MQ-20 Avenger as a flying laboratory for this purpose. The aircraft is a high-speed, multi-mission unmanned aircraft system designed to operate in high-threat environments. Its stealth-focused design and S-shaped exhaust reduce both radar visibility and heat signature, making it suitable for air combat trials.

For this latest test, the MQ-20 acted as a stand-in for a future Collaborative Combat Aircraft. Its task was to intercept and engage a human-piloted aggressor aircraft. The company has not identified the opposing aircraft, though earlier tests suggest it may have been similar to an F-5 Tiger II.

Sensors and software working without radar

During the exercise, the MQ-20 relied on a live Anduril Infrared Search and Track sensor. Unlike radar, this system detects heat rather than emitting signals. That allowed the drone to locate and follow the aggressor aircraft without revealing its own position.

Once the target was detected, the MQ-20 used onboard computers to build a track file. The software predicted the target’s flight path and calculated an intercept solution. From there, the system generated a firing solution and executed a simulated weapon shot. Telemetry data from the exercise confirmed the outcome as a successful kill.

This sequence demonstrated how sensor data and autonomy can work together. The drone did not rely on constant human input. Instead, it processed information, made decisions, and carried out the mission using its own logic while operating across miles of contested airspace.

Flying aggressively while respecting airspace rules

A critical part of the exercise was not just the intercept but also how the MQ-20 behaved during it. Even during aggressive maneuvering, the drone followed standard rules for operating in integrated airspace. This is essential for future operations where military and civilian aircraft may share the sky.

The MQ-20 stayed within its assigned Keep In Zone and avoided all Keep Out Zones. These boundaries are defined in miles and feet to prevent aircraft from entering restricted civilian areas or dangerous threat corridors. By respecting these limits, the drone proved it could fight without becoming a hazard to others.

This balance between aggression and control is a key requirement for autonomous combat systems. It shows that autonomy does not mean unpredictability.

Switching autonomy modes without disruption

The test also highlighted the MQ-20’s ability to manage flight and mission tasks smoothly. The aircraft executed standard instrument flying, maintaining heading, speed, and altitude while avoiding terrain. It transitioned between flight autonomy and mission autonomy without any disruption.

This flexibility is essential for future combat drones that must adapt to changing conditions. A system that can handle basic flying tasks while also managing complex combat decisions reduces the burden on human operators.

“This demonstration reinforces our commitment to advancing Human-Machine Teaming and highlights the growing sophistication of autonomous systems in using sensor data to make independent decisions,” said Michael Atwood, Vice President of Advanced Programs at General Atomics.

“The ability of autonomy to close on a target using its own logic is a vital step toward building a reliable ecosystem of collaborative combat aircraft for the modern warfighter.”

🔗 Sumber: interestingengineering.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