MAROKO133 Hot ai: Harvard’s 448-qubit breakthrough brings fault-tolerant quantum supercomp

📌 MAROKO133 Update ai: Harvard’s 448-qubit breakthrough brings fault-tolerant quan

Researchers at Harvard, publishing in the journal Nature on Monday, have demonstrated a new quantum system that can detect and correct errors below a critical threshold. 

The breakthrough offers a potential solution to quantum error correction, which has long been considered the single greatest obstacle to building practical, large-scale quantum computers.

The central challenge has been that “qubits”—the basic units of quantum information—are inherently fragile. 

Unlike the stable “bits” (ones and zeros) in conventional computers, qubits are subatomic particles that are highly susceptible to slipping out of their quantum states and losing their encoded information due to environmental noise.

This fragility has been a formidable roadblock because the very properties that make quantum computers so powerful also make them prone to errors. These machines promise to leverage quantum phenomena, such as entanglement, to achieve processing power that grows exponentially. 

“In theory, a system of 300 quantum bits can store more information than the number of particles in the known universe,” said the researchers in a press release.

A fault-tolerant system

To overcome the error hurdle, the Harvard-led team built a “fault-tolerant” system using 448 atomic qubits. The system, which specializes in using neutral atoms of the element rubidium, employs an intricate sequence of techniques, including physical and logical entanglement, logical magic, and even “quantum teleportation”—the transfer of a quantum state from one particle to another without physical contact.

This new architecture is the first to suppress errors below the critical point, where adding more qubits to the system further reduces errors rather than increasing them, a requirement for scaling up the technology.

“For the first time, we combined all essential elements for a scalable, error-corrected quantum computation in an integrated architecture,” said Mikhail Lukin, co-director of the Quantum Science and Engineering Initiative at Harvard and the senior author of the paper.

Dolev Bluvstein, the paper’s lead author and now an assistant professor at Caltech, emphasized the system’s scalability. 

“There are still a lot of technical challenges remaining to get to a very large-scale computer with millions of qubits, but this is the first time we have an architecture that is conceptually scalable,” stated Bluvstein. 

“It’s becoming clear that we can build fault-tolerant quantum computers.”

A series of advances

The research was a collaboration between Harvard and MIT, jointly headed by Lukin, Markus Greiner of Harvard, and Vladan Vuletić of MIT. The team also collaborates with researchers at QuEra Computing, the Joint Quantum Institute at the University of Maryland, and the National Institute of Standards and Technology.

This success is the latest in a series of advances from the group. In September, they published another Nature paper demonstrating a system with over 3,000 qubits that could operate for more than two hours, solving a different technical hurdle related to atom loss.

Lukin believes the long-held goal of functional quantum computing is finally becoming a reality.

“This big dream that many of us had for several decades, for the first time, is really in direct sight,” he said.

🔗 Sumber: interestingengineering.com

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

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