# Jun 14, 2025

## Self-Adapting Language Models

- **Self-Adapting LLMs (SEAL)** enable large language models to **generate their own finetuning data** and update directives, allowing for dynamic adaptation to new tasks and knowledge.

## Seven replies to the viral Apple reasoning paper – and why they fall short

- The **Apple paper** critiques the **limitations of Large Reasoning Models (LRMs)**, asserting they often fail to execute complex tasks reliably, highlighting a significant gap in the pursuit of Artificial General Intelligence (AGI).

## Saab achieves AI milestone with Gripen E

- **Saab's Gripen E** has successfully integrated Helsing's AI agent, **Centaur**, completing three flights that demonstrate its ability to autonomously execute complex maneuvers in combat scenarios, marking a significant leap in military aviation technology.

## [P] 3Blue1Brown Follow-up: From Hypothetical Examples to LLM Circuit Visualization

- **Significant advancements** in mechanistic interpretability now allow us to "decompose" LLMs into interpretable circuits, enhancing our understanding of how models generate predictions from input sentences.

## The Tech Job Meltdown

- **Over half a million tech layoffs** stem from the 2022 tax reform that mandated R&D costs be capitalized and amortized, drastically increasing short-term tax liabilities for companies reliant on innovation.

## Unsupervised Elicitation of Language Models

- The **Internal Coherence Maximization (ICM)** algorithm fine-tunes pretrained language models using their own generated labels, achieving performance comparable to **golden supervision** and surpassing **crowdsourced human supervision** on various tasks. [Link to article](https://arxiv.org/abs/2506.10139)

## [2506.06105] Text-to-LoRA: Instant Transformer Adaption

- **Text-to-LoRA (T2L)** enables **instant adaptation** of large language models (LLMs) using only a natural language description, eliminating the need for extensive dataset curation and fine-tuning.

## How to Build Conscious Machines

- **OSF** provides a platform that requires **JavaScript** for full functionality, emphasizing the need for users to enable it for optimal use.

## Writing a Truth Oracle in Lisp

- The **truth oracle** in Lisp aims to determine the truth value of arbitrary mathematical statements, leveraging the **Curry-Howard correspondence** to connect formal logic proofs with typed functional programming constructs, particularly in typed Racket.

## [D] Nvidia’s “Join Us or Compete” moment — the GPU cloud stack is collapsing

- **Nvidia is evolving** from a chip manufacturer to a comprehensive **AI infrastructure provider**, offering full servers, APIs, and inference microservices, fundamentally altering the competitive landscape in the GPU cloud market.

## We investigated Amsterdam's attempt to build a 'fair' fraud detection model

- **Amsterdam's fraud detection model aimed to reduce investigations while ensuring fairness, yet initial tests revealed significant bias against non-Dutch applicants, with a 30% higher false positive rate.** This model, developed using an Explainable Boosting Machine, sought to balance efficiency and ethical considerations in welfare fraud detection.

## [D][R] Collaborative Learning in Agentic Systems: A Collective AI is Greater Than the Sum of Its Parts

- **MOSAIC** is a groundbreaking framework that enables **collaborative learning** among autonomous AI agents, allowing them to share and reuse knowledge without centralized control, enhancing both **speed** and **performance** in dynamic environments.

## Peeling the Covers Off Germany's Exascale "Jupiter" Supercomputer

- **Jupiter**, Germany's first exascale supercomputer, is a hybrid **CPU-GPU** system primarily utilizing **Nvidia** technology, highlighting Europe's ongoing struggle for chip independence despite initial plans for custom hardware.

## [D] Geometric NLP

- **Geometric NLP** explores **non-Euclidean geometry** to enhance our understanding of natural language, revealing insights into **embedding landscapes** and concept hierarchies that can improve AI safety and mechanistic interpretability.

## [P] Residual Isolation Forest

- The **Residual Isolation Forest** is a novel estimator designed for **contextual anomaly detection**, enhancing Isolation Forest performance by leveraging **residuals** from semantically separated variable groups.