KévinInteresting take. Worth the read if you have time.
Hacker NewsA deep dive into self-improving AI and the Darwin-Gödel Machine
https://richardcsuwandi.github.io/blog/2025/dgm/
#HackerNews #selfimprovingAI #DarwinGödelMachine #AIresearch #technology #innovation
Deep learning gets the glory, deep fact checking gets ignored
https://rachel.fast.ai/posts/2025-06-04-enzyme-ml-fails/index.html
#HackerNews #deeplearning #factchecking #ML #ethics #AIresearch
ReasoningGym: Reasoning Environments for RL with Verifiable Rewards
https://arxiv.org/abs/2505.24760
#HackerNews #ReasoningGym #ReinforcementLearning #VerifiableRewards #AIResearch #MachineLearning
REASONING GYM: Reasoning Environments for Reinforcement Learning with Verifiable Rewards
We introduce Reasoning Gym (RG), a library of reasoning environments for reinforcement learning with verifiable rewards. It provides over 100 data generators and verifiers spanning multiple domains including algebra, arithmetic, computation, cognition, geometry, graph theory, logic, and various common games. Its key innovation is the ability to generate virtually infinite training data with adjustable complexity, unlike most previous reasoning datasets, which are typically fixed. This procedural generation approach allows for continuous evaluation across varying difficulty levels. Our experimental results demonstrate the efficacy of RG in both evaluating and reinforcement learning of reasoning models.
SRF IRIS🤝 In collaboration with the (@UniStuttgartAI) Institute for Artificial Intelligence at the Universität Stuttgart, it is our great pleasure to highlight the following event:
🎤 Engineering Safe Systems with AI
🗓️ June 5, 2025 | 15:45 | Room U32.101, Universitätsstraße 32
We’re pleased to support this talk by Dr. Reinhard Stolle Deputy Director at Fraunhofer IKS, on how to engineer safe AI-enabled systems without compromising innovation.
In his talk, “Engineering Safe Systems with AI”, Dr. Stolle will explore two key perspectives on safety: a safety-centricand an AI-centric view. He will present his team’s approach to combining the strengths of both, introducing a model for continuous safety engineering for high-risk AI systems—explicitly modeling and propagating uncertainties and confidences during both design and operation.
📣 Students, staff, and all interested guests are warmly invited to attend this exciting and insightful session!
👤 About the Speaker
Dr. Reinhard Stolle is Deputy Director of Fraunhofer IKS and Head of the Mobility Business Unit. He studied computer science at FAU Erlangen and the University of Colorado at Boulder, earned his master’s and Ph.D. in AI, and completed postdoctoral research at Stanford. His career spans AI research at Xerox PARC, 14 years in software and autonomous driving at BMW, and leadership roles at AID (VW Group) and Argo AI, focusing on Level 4 autonomous vehicles.
#AI
#SafeAI
#Engineering
#FraunhoferIKS
#AIsafety
#AutonomousSystems
#TechTalk
#Innovation
#ContinuousEngineering
#AIethics
#KIInstitut
#AIresearch
IRIS Board of Directors
Prof. Dr. André Bächtiger
Prof. Dr. Reinhold Bauer
Prof. Dr. Sibylle Baumbach
Dr. Miriam K.
Prof. Dr. Steffen Staab @ai
Jun.-Prof. Dr. Maria Wirzberger
N-gated Hacker NewsApparently, AI researchers have stumbled upon the groundbreaking discovery that going backward is the way forward in saving energy. 🚶♂️🔙 Who knew the secret to progress was simply hitting rewind? 🤦♂️ Next, they'll suggest ditching computers for abacuses to save on electricity. 🧮💡
https://www.quantamagazine.org/how-can-ai-researchers-save-energy-by-going-backward-20250530/ #AIresearch #energyconservation #innovation #backwardthinking #technologyhumor #HackerNews #ngated
https://www.quantamagazine.org/how-can-ai-researchers-save-energy-by-going-backward-20250530/ #AIresearch #energyconservation #innovation #backwardthinking #technologyhumor #HackerNews #ngated
froggfishacademia.edu gebruikte een van mijn presentaties om een AI podcast te maken, zonder mij te vragen. Vraagt om een waardering. Wanneer ik eerst wil luisteren, werkt link niet. Wordt gedwongen om te betalen om naar mijn "eigen" podcast te luisteren!!
https://www.academia.edu/ai_podcast/5360998
Does the fediverse have a similar place as academia.edu?
#opensource #podcast #AI #AIresearch #https://mas.to/@nlnet@nlnet.nl
froggfishacademia.edu used one of presentations to realise an AI podcast of my presentation without asking me. They want me to rate it. But when I want to listen to the podcast of my presentation the link doesnt work. There is a button to take a subscription. Do they force me to take a subscription to listen to a podcast of my own presentation?
Does the fediverse have a similar place as academia.com?
#opensource #podcast #AI #AIresearch #https://mas.to/@fsf@hostux.social #https://mas.to/@nlnet@nlnet.nl
Beyond Attention: Toward Machines with Intrinsic Higher Mental States
https://arxiv.org/abs/2505.06257
#HackerNews #BeyondAttention #IntrinsicMentalStates #AIResearch #MachineLearning #CognitiveScience
Beyond Attention: Toward Machines with Intrinsic Higher Mental States
Attending to what is relevant is fundamental to both the mammalian brain and modern machine learning models such as Transformers. Yet, determining relevance remains a core challenge, traditionally offloaded to learning algorithms like backpropagation. Inspired by recent cellular neurobiological evidence linking neocortical pyramidal cells to distinct mental states, this work shows how models (e.g., Transformers) can emulate high-level perceptual processing and awake thought (imagination) states to pre-select relevant information before applying attention. Triadic neuronal-level modulation loops among questions ($Q$), clues (keys, $K$), and hypotheses (values, $V$) enable diverse, deep, parallel reasoning chains at the representation level and allow a rapid shift from initial biases to refined understanding. This leads to orders-of-magnitude faster learning with significantly reduced computational demand (e.g., fewer heads, layers, and tokens), at an approximate cost of $\mathcal{O}(N)$, where $N$ is the number of input tokens. Results span reinforcement learning (e.g., CarRacing in a high-dimensional visual setup), computer vision, and natural language question answering.
YOLO-World: Real-Time Open-Vocabulary Object Detection
https://arxiv.org/abs/2401.17270
#HackerNews #YOLO #World #RealTime #ObjectDetection #OpenVocabulary #AIResearch
YOLO-World: Real-Time Open-Vocabulary Object Detection
The You Only Look Once (YOLO) series of detectors have established themselves as efficient and practical tools. However, their reliance on predefined and trained object categories limits their applicability in open scenarios. Addressing this limitation, we introduce YOLO-World, an innovative approach that enhances YOLO with open-vocabulary detection capabilities through vision-language modeling and pre-training on large-scale datasets. Specifically, we propose a new Re-parameterizable Vision-Language Path Aggregation Network (RepVL-PAN) and region-text contrastive loss to facilitate the interaction between visual and linguistic information. Our method excels in detecting a wide range of objects in a zero-shot manner with high efficiency. On the challenging LVIS dataset, YOLO-World achieves 35.4 AP with 52.0 FPS on V100, which outperforms many state-of-the-art methods in terms of both accuracy and speed. Furthermore, the fine-tuned YOLO-World achieves remarkable performance on several downstream tasks, including object detection and open-vocabulary instance segmentation.