📌 TOPINDIATOURS Update ai: Professor Says Her Garbled AI Textbook Was a Huge Succe

The professor behind an AI-generated textbook says that her error-ridden experiment was actually a resounding success.

Designed for a comparative literature course on medieval and Renaissance-era writing and announced by UCLA at the end of 2024, the digital textbook was immediately met with widespread mockery and derision from educators. Its AI-generated cover was riddled with incomprehensible text — “Of Nerniacular Latin To An Evoolitun On Nance Langusages,” for example — and featured generic visuals that had little to do with the period it was supposedly covering. 

At the time, Elizabeth Landers, a grad student who helped put together the volume, said that the errors “aren’t a failure of AI.” Instead, she argued, “they’re an intentional artistic choice that prompts students to question their assumptions about language, meaning and historical truth.”

Now in a new interview with Inside Higher Ed in which the word “hallucination” isn’t mentioned once, the course’s professor Zrinka Stahulja called her decision to use an “AI-assisted” textbook a “no-brainer” because of all the time it saved her, helping her be an “approachable and accessible teacher.”

And incredibly, Stahuljak says she was surprised that her UCLA colleagues were so skeptical about her AI textbook. “I was really shocked that they couldn’t see that this textbook was my creation; it was carefully edited, just as if it had been printed,” she told IHE.

“I don’t see how a traditional textbook that costs $250 and is out of date within two years or three years, would be in some way better than a custom $25 AI-facilitated textbook that is based on my material,” she added.

The AI textbook was made with Kudu, a platform for creating digital textbooks started by another UCLA professor. Stahuljak says she created the textbook by supplying her own notes to the AI tool, which was instructed not to pull from outside sources. Students could interact with a built-in chatbot to help learn the materials, though she stresses it was designed not to write papers or complete assignments. Stahuljak also says the AI features made the book more accessible, with some students saying they listened to it while walking or at the gym.

After deploying the AI textbook, Stahuljak claimed that “engagement went up” compared to classes that didn’t use it. And perhaps soberingly, she viewed it as a preferable to having her students turn to ChatGPT for help.

“It’s better than some commercial version that has nothing to do with what you’re teaching or is pulling the information from the internet,” she said in the interview. “We’re losing that control when we are indiscriminately given ChatGPT or other commercial generative AI-powered tools.”

There’s a fair point or two being made, but Stahuljak isn’t addressing the numerous elephants in the room. AI chatbots are notorious for generating made-up facts and otherwise incorrectly reporting information, regardless of whatever data they’re being asked to pull from. A considerable and still growing body of evidence shows how AI tools may diminish critical thinking skills and attention spans. Then there’s the broader concerns over how the tech is threatening the very existence of learning institutions, as tech companies spend millions of dollars to capture schools and universities and use them to offload their products.

“This is truly bad and makes me wonder if we aren’t participating in creating our own replacements at the expense of, well, everyone who cares about teaching and learning,” one English professor wrote on social media after the AI textbook was announced, as quoted by IHE.

Others were even harsher.

“If you do this you should have your doctorate revoked and be thrown into the stocks at the center of the main university quad,” fumed another professor. “This is abandonment of professional responsibility to a degree that would be comical if it weren’t so self-serious.”

More on AI: Gen Z Arriving at College Unable to Read

The post Professor Says Her Garbled AI Textbook Was a Huge Success 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