Larry Ellison wants to put all America's data in AI, including DNA

Larry Ellison, the billionaire founder of Oracle, wants to create a massive database system that stores all of America's data, including DNA information, to be analyzed by AI. He believes this would lead to improvements in healthcare, food production, and other areas, but it also raises concerns about privacy and surveillance. Ellison's proposal is part of his vision for the future of AI, which he thinks will revolutionize various aspects of life, and he sees Oracle as a key player in making this vision a reality.

Read more

OpenAI's GPT-5 Bait and Switch

OpenAI's announced plans to merge its O-series and GPT models into a unified system, eliminating model selection and replacing it with "intelligence levels" tied to subscription tiers, which may reduce user control and transparency. This shift, framed as "magic unified intelligence," could negatively impact power users who rely on specific model capabilities, and may ultimately result in a cost-optimized version of GPT-5 that reserves advanced capabilities for internal use or limited applications.

Read more

The demise of software engineers due to AI is greatly exaggerated

Senior managers have high expectations for AI replacing software engineers, but the reality is that AI is not yet capable of fully replacing them, and is currently being used by engineers as a tool to increase productivity, such as with code auto-completion. While AI may be able to assist with certain tasks, it is not ready to handle the complex and varied responsibilities of a software engineer, and hiring of engineers is likely to continue.

Read more

Evaluating LLM Reasoning Through Live Computer Games

The Roblox Game Arena offers several escape games where players can compete against AI, including Akinator, Taboo, and Bluffing. The games feature leaderboards that track player performance, with an overall ranking calculated based on average scores across all games.

Read more

AI summaries turn real news into nonsense, BBC finds

The BBC has conducted research on the accuracy of AI-powered news summaries, finding that 51% of responses from popular AI assistants, including ChatGPT and Google's Gemini, contained significant issues such as factual errors, misquotes, and lack of context. The study revealed that these AI assistants often introduced inaccuracies, even when citing BBC news articles as sources, highlighting the need for responsible use of AI in news summarization.

Read more

Data Formulator 2: Iteratively Creating Rich Visualizations with AI

Data analysts often struggle with iterative visualization authoring, requiring proficiency in data transformation and visualization tools, as well as managing multiple versions of data and charts. Data Formulator 2, an LLM-powered visualization system, addresses these challenges by allowing users to describe their visualization intent through blended UI and natural language inputs, and supports iteration by enabling users to navigate and reuse previous designs.

ArXiv

Decoding AI Judgment: How LLMs Assess News Credibility and Bias

This study examines how Large Language Models (LLMs) assess news credibility by benchmarking their reliability and political classifications against expert-driven rating systems and analyzing the linguistic markers that shape their decisions. The research also introduces a framework to refine LLM credibility assessments by incorporating external information and querying other models to determine if their evaluations are based on structured reasoning or prior learned associations.

ArXiv

LM2: Large Memory Models

The Large Memory Model (LM2) is a decoder-only Transformer architecture that incorporates an auxiliary memory module to improve performance in multi-step reasoning and long-context tasks. Experimental results show that LM2 outperforms existing models, achieving significant improvements on benchmarks such as BABILong and MMLU, while maintaining general-purpose capabilities and demonstrating the importance of explicit memory in enhancing Transformer architectures.

ArXiv

Magnetic field sorting of superconducting graphite particles with Tc>400K (2024)

Researchers have developed a method to separate superconducting particles from normal grains in industrial graphite powders using magnetic field gradients, resulting in a concentrate of particles that exhibit superconductivity above room temperature. Electrical resistance and magnetization measurements confirm the superconducting properties, with transition temperatures reaching up to 700K and zero resistance up to 500K, paving the way for further study of these unique graphite phases.

ArXiv

Vision-Language Models vs. Traditional OCR in Video – New Benchmark

This paper introduces a benchmark for evaluating Vision-Language Models (VLMs) on Optical Character Recognition (OCR) tasks in dynamic video environments, using a curated dataset of 1,477 annotated frames. The benchmark highlights the potential of VLMs to outperform traditional OCR systems, but also reveals challenges such as hallucinations and sensitivity to occluded text, with the dataset and framework made publicly available for further research.

ArXiv

Show HN: Dockershrink – AI Assistant to reduce the size of Docker images

DockerShrink is an AI-powered command-line tool that helps reduce the size of Docker images by applying state-of-the-art optimizations, including multi-stage builds and lighter base images. The tool, currently in beta and supporting only NodeJS applications, can generate optimized Docker image definitions and optimize existing ones, aiming to save storage and transfer costs, decrease build times, and increase developer productivity.

Repo

Show HN: Cognee – Turn RAG and GraphRAG into custom dynamic semantic memory

Cognee is a data layer for AI applications that implements scalable, modular ECL (Extract, Cognify, Load) pipelines, allowing developers to interconnect and retrieve past conversations, documents, and audio transcriptions while reducing hallucinations, effort, and cost. It merges graph and vector databases to uncover hidden relationships and patterns in data, and can be installed using pip or poetry, with support for various databases and vector stores.

Repo

Show HN: Morph – Open-Source Python and Markdown Framework for AI Apps

Morph is a Python-centric full-stack framework for building and deploying data apps, allowing users to get started quickly with just three commands and create visually appealing designs without needing HTML/CSS knowledge. The framework supports customizable data workflows with Python and SQL, and provides a simple development process involving creating Python and SQL files and connecting them to pages defined in MDX files.

Repo

I built an AI aimbot thats in runs in the browser

There is no text to summarize. The input appears to be an error message indicating that a README file could not be retrieved.

Repo

Show HN: Hardware-Friendly Text Classification with Model2Vec

Model2Vec is a technique that turns sentence transformers into small, fast, and powerful static models, reducing model size by up to 50 times and making them up to 500 times faster with minimal performance drop. The Model2Vec library provides pre-trained models, distillation capabilities, and fine-tuning options, allowing users to easily integrate static embeddings into their applications.

Repo