The day's stories
FundingMeta's loss is Thinking Machines' gainIn a notable shift within the tech industry, Weiyao Wang, a former employee of Meta, has joined Thinking Machines Lab (TML) as the startup gears up for expansioTechCrunch AIApr 24· 59%FundingGoogle to invest up to $40B in Anthropic in cash and computeGoogle has announced a significant investment of up to $40 billion in Anthropic, a prominent AI research company. The initial phase of this investment includes TechCrunch AIApr 24· 59%OtherDeepSeek-V4: a million-token context that agents can actually useDeepSeek has unveiled its latest model, DeepSeek-V4, which boasts an impressive capability of handling a million-token context window. This significant upgrade Hugging Face BlogApr 24· 59%FundingComfyUI hits $500M valuation as creators seek more control over AI-generated mediaComfyUI, a startup that emerged from an open-source initiative, has recently achieved a significant milestone by raising $30 million, bringing its valuation to TechCrunch AIApr 24· 59%OtherDeepSeek previews new AI model that 'closes the gap' with frontier modelsDeepSeek has introduced its latest advancements in AI with the unveiling of two new large language models, DeepSeek V4 Flash and V4 Pro. These models incorporatTechCrunch AIApr 24· 59%InfrastructureIn another wild turn for AI chips, Meta signs deal for millions of Amazon AI CPUsMeta has recently entered into a significant agreement with Amazon to utilize millions of AWS Graviton chips for its artificial intelligence workloads. This parTechCrunch AIApr 24· 59%OtherMarked-up Mac minis flood eBay amid shortages driven by AIThe recent surge in demand for Mac minis, especially the M4 model, has resulted in significant shortages, leading to inflated prices on platforms like eBay. ThiTechCrunch AIApr 24· 59%Model ReleaseDeepSeek's new models offer big inference cost savingsDeepSeek has unveiled its latest V4 models, which are designed to significantly cut inference costs while maintaining competitive performance against leading prThe Register AIApr 24· 59%OtherIntel expects AI inference to drive demand for its CPUsIntel's CEO, Lip-Bu Tan, has expressed confidence that AI inference will significantly boost demand for the company's CPUs, marking a strategic pivot in their bThe Register AIApr 24· 59%Product LaunchNothing introduces an AI-powered dictation toolNothing has unveiled a new AI-powered dictation tool named Essential Voice, designed to enhance the transcription experience by converting speech into well-struTechCrunch AIApr 24· 58%Open SourceOpen source models can find bugs as well as MythosAt Black Hat Asia, Ari Herbert-Voss, CEO of RunSybil and OpenAI's first security hire, highlighted the capabilities of open source AI models in identifying softThe Register AIApr 24· 57%OtherMeta to use millions of AWS Graviton coresMeta has entered into a multi-year agreement to leverage tens of millions of AWS Graviton 5 CPU cores, marking a pivotal transition towards Arm-based processorsThe Register AIApr 24· 57%Product LaunchAutomated Detection of Dosing Errors in Clinical Trial Narratives: A Multi-Modal Feature Engineering Approach with LightGBMRecent advancements in the detection of dosing errors within clinical trial narratives have been made through the development of an automated system utilizing LarXiv cs.AIApr 24· 57%OtherFairyFuse: Multiplication-Free LLM Inference on CPUs via Fused Ternary KernelsFairyFuse is a groundbreaking inference system aimed at optimizing the performance of large language models (LLMs) on CPUs. By leveraging fused ternary kernels,arXiv cs.LGApr 24· 57%OtherEvoForest: A Novel Machine-Learning Paradigm via Open-Ended Evolution of Computational GraphsEvoForest has emerged as a groundbreaking hybrid neuro-symbolic system that innovatively evolves computational graphs to push the boundaries of machine learningarXiv cs.AIApr 24· 57%ResearchAbsorber LLM: Harnessing Causal Synchronization for Test-Time TrainingThe introduction of the Absorber LLM marks a significant advancement in the efficiency of transformer models during inference, particularly for long sequences. arXiv cs.LGApr 24· 57%OtherReinforcing privacy reasoning in LLMs via normative simulacra from fictionRecent research has introduced an innovative approach to enhance privacy reasoning in large language models (LLMs) by leveraging normative simulacra from fictioarXiv cs.LGApr 24· 57%SafetyInference Headroom Ratio: A Diagnostic and Control Framework for Inference Stability Under ConstraintRecent research has introduced the Inference Headroom Ratio (IHR), a novel diagnostic tool aimed at enhancing inference stability in constrained decision-makingarXiv cs.AIApr 24· 56%BenchmarkExplainable AML Triage with LLMs: Evidence Retrieval and Counterfactual ChecksRecent advancements in anti-money laundering (AML) practices have seen the introduction of a novel framework that leverages large language models (LLMs) for exparXiv cs.AIApr 24· 56%OtherExploring Data Augmentation and Resampling Strategies for Transformer-Based Models to Address Class Imbalance in AI Scoring of Scientific Explanations in NGSS ClassroomRecent research has focused on enhancing the performance of transformer-based models in the automated scoring of scientific explanations within the Next GeneratarXiv cs.AIApr 24· 56%OtherILDR: Geometric Early Detection of GrokkingRecent research has unveiled a novel metric known as the Inter/Intra-class Distance Ratio (ILDR), designed to facilitate the early detection of the grokking phearXiv cs.LGApr 24· 56%Product LaunchThe Tool-Overuse Illusion: Why Does LLM Prefer External Tools over Internal Knowledge?Recent research has delved into the tendency of large language models (LLMs) to excessively rely on external tools rather than utilizing their internal knowledgarXiv cs.AIApr 24· 56%OtherAlgorithm Selection with Zero Domain Knowledge via Text EmbeddingsRecent advancements in algorithm selection have been showcased through the introduction of ZeroFolio, a novel method that leverages pretrained text embeddings tarXiv cs.AIApr 24· 56%OtherFrom Actions to Understanding: Conformal Interpretability of Temporal Concepts in LLM AgentsRecent research has introduced a novel framework aimed at enhancing the interpretability of temporal concepts in Large Language Models (LLMs) through the use ofarXiv cs.AIApr 24· 56%ResearchDo Masked Autoencoders Improve Downhole Prediction? An Empirical Study on Real Well Drilling DataRecent research has focused on the application of masked autoencoders (MAE) to enhance the accuracy of downhole drilling predictions using real telemetry data. arXiv cs.LGApr 24· 56%RegulationAI to Learn 2.0: A Deliverable-Oriented Governance Framework and Maturity Rubric for Opaque AI in Learning-Intensive DomainsThe recent paper titled 'AI to Learn 2.0' presents a governance framework aimed at enhancing the use of generative AI in learning-intensive domains. This framewarXiv cs.AIApr 24· 56%OtherClinically Interpretable Sepsis Early Warning via LLM-Guided Simulation of Temporal Physiological DynamicsRecent research has introduced a groundbreaking framework that leverages Large Language Models (LLMs) to enhance the early warning systems for sepsis, a life-tharXiv cs.LGApr 24· 56%FundingThermoQA: A Three-Tier Benchmark for Evaluating Thermodynamic Reasoning in Large Language ModelsResearchers have unveiled ThermoQA, a novel benchmark consisting of 293 thermodynamics problems designed to evaluate the reasoning capabilities of large languagarXiv cs.AIApr 24· 56%OtherValidating a Deep Learning Algorithm to Identify Patients with Glaucoma using Systemic Electronic Health RecordsRecent research has successfully validated a deep learning algorithm designed to identify patients at high risk for glaucoma by utilizing systemic electronic hearXiv cs.LGApr 24· 55%