Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
A Pew Research Center poll revealed that 26% of teens now use ChatGPT for homework assistance, doubling last year's figure. Comfort with AI use varies, with 54% accepting it for research but only 18% for essay writing. While initial bans were imposed, many school districts have eased restrictions on classroom AI use.
EvaByte is a model from SambaNova trained on byte level language modeling. It performs on par with models two generations old (e.g., Llama 2). While this may seem slightly disappointing, it is a big win for byte level models, which have been stuck at sub 1B parameter scale for some time.
A novel approach to improved sample efficiency for Diffusion models that doesn't involve modifications to the underlying architecture.
A comprehensive review of machine learning applications in archaeological practices.
MedSSS is a slow-thinking small medical language model built with a self-evolution pipeline.
MTU-Bench is a new benchmark designed to evaluate large language models' tool-use capabilities across various scenarios.
MMAudio is a video-to-audio system that takes videos as input and creates paired audio based on the video's content. It works well with both synthetic and real videos.
Google DeepMind is forming a team led by Tim Brooks to develop AI "world models" for simulating physical environments. These models aim to enhance real-time interactive media and training scenarios, contributing to Google's pursuit of AGI. The initiative will collaborate with existing Google AI projects like Gemini and Veo.
AMR-Policy GPT is an AI tool that uses LLMs to aid in antimicrobial resistance policy development across 146 countries. The tool provides policymakers with contextually relevant information, fostering better-informed National Action Plans, especially in regions lacking local data or infrastructure. The project is co-led by academics from the Chinese Academy of Sciences and Durham University.
AI predictions for 2025 include advancements in customized agents, generative video, and versatile robots. Meta is dropping professional fact-checkers, shifting responsibility to users, causing concern among its former partners. MIT Technology Review has highlighted emerging stem-cell therapies as one of ten breakthrough technologies for 2025.
Sponsored
RAG in a few lines of code? Writer's predefined RAG tool integrates your data into a Knowledge Graph, enabling AI interactions with a single API call. No complex setups, just scalable, accurate AI
Heyboss is an AI engineer for non-coders that can build websites, apps, and games.
Nvidia CEO Jensen Huang predicts that AI robots will first impact manufacturing, boosting productivity and offsetting lost revenues, thus reducing inflation.
Blackstone is investing $300 million in data storage firm DDN, valuing it at $5 billion.
Sponsored
The Generative AI revolution has reached its reckoning. In 2025, leaders face new challenges: Prove ROI, harness AI agents, and transform governance into a strategic asset, or risk falling behind. This report explores five critical shifts [4]:
Mistral, a French AI startup, plans to take the initial public offering route instead of being acquired. It has raised substantial capital and is well positioned to continue providing service globally, and specifically to European companies.
NVIDIA presents a novel approach that combines foundation models with stereo matching techniques to enhance 3D perception for robots and autonomous vehicles.
DeepSeek is one of the only open model providers that seems to be innovating solidly on the underlying architecture of its generative models. Specifically, its research improvements are not the result of brute force, but instead come from minds that clearly understand what the Transformer does and operate in that paradigm.
This paper introduces "landmarker," a Python package built on PyTorch that offers a flexible toolkit for developing and evaluating landmark localization algorithms.
UniAct is a new embodied foundation modeling framework that operates in the Universal Action Space.
The new Dolphin model, built on Llama 3.1 8B, was trained to take instructions via the system prompt. This means that it is exceptionally good at various personas and roleplay while maintaining much of the power of the base Llama model itself.
Sponsored
Get in front of your target audience right here in TLDR! No bid adjustments, algorithms to fight, or campaign and targeting management
SurgeNet is a novel surgical foundation model that sets a new benchmark in surgical computer vision.
This project offers an intuitive web interface to make tracking workouts easier and more effective.
DETRIS enhances visual feature propagation for misaligned encoders in multimodal tasks that employs dense interconnections and text adapters.
AI query engines enable enterprises to effectively utilize vast amounts of both structured and unstructured data, bridging the gap between raw data and AI-powered applications. They offer advanced features like diverse data handling, scalability, accurate retrieval, and continuous learning, enhancing the capabilities of AI agents. Companies like DataStax are already leveraging these engines to support applications in customer service, video search, and software analysis.
Sam Altman discusses his unexpected firing from OpenAI's board and reflects on its impact on governance and his leadership. He addresses critiques about OpenAI's strategy, emphasizing the company's goal to safely iterate towards AGI, while acknowledging high-profile departures. Altman shares his thoughts on the potential risks of AGI and hints at AI workforce integration by 2025, asserting that OpenAI remains focused on building superintelligent systems for future prosperity.
Stanford researchers created AI agents that replicate human behavior with 85% accuracy by interviewing 1,052 people using a purpose-built interface. The study, in collaboration with Google DeepMind, suggests these agents could help decision-makers understand public opinions by simulating responses to policies and product launches. While AI agents performed well on personality tests, they struggled with replicating human decisions in economic games.
Sponsored
Aisera's Agentic AI products help enterprises accelerate revenue revenue, increase productivity, and deliver exceptional user experiences. See how Gartner views this market. Get the Magic Quadrant
Inworld AI, Streamlabs, and Nvidia are launching an AI assistant for streamers to provide real-time technical support and act as a game-aware cohost.
Microsoft has released its Phi-4 model as open-source on Hugging Face, with the model praised for its efficiency and high performance in reasoning tasks.
AI can help stroke and Parkinson's patients communicate by interpreting throat vibrations.
Sponsored
Does your app get fake signups, throwaway emails, or users abusing your free tier? Or worse, bots attacks and and brute force attempts?
DeepSeek says that its latest reasoning-focused AI model outperforms OpenAI's O1 model on specific benchmarks, citing improvements in logical inference and factual accuracy.
OpenAI's record success with the o3 model on the FrontierMath benchmark revealed its undisclosed funding of the project. Despite excellence in solving complex math problems, a lapse in transparency was admitted by benchmark developer Epoch AI. Future collaboration promises improved openness, particularly regarding funding and data access.
Perplexity AI has submitted a bid to merge with TikTok US as a TikTok ban approaches. The proposal aims to combine Perplexity, TikTok US, and new equity partners while allowing ByteDance investors to retain their equity. Perplexity's merger offer seeks to integrate more video content into its AI search operations.
OpticFusion is a multimodal neural implicit 3D reconstruction of microstructures that fuses white light interferometry and optical microscopy.
DeepSeek has released a paper along with its strong MIT licensed reasoning models. Importantly, the paper shows that there is no need for extremely fancy process reward models. DeepSeek simply used accuracy of answer, a formatting reward, and large scale RL to encourage reasoning in the model. It found that distilling this reasoning ability via SFT to smaller models gave an incredible boost in final performance.
Contrary to the direct R1 approach, this work from DeepMind uses an evolutionary strategy, guided by a language model, to modify and propose new responses. This leads to strong reasoning ability, specifically in planning tasks that require some exploration.
Sponsored
In 2025, businesses will increasingly use agentic AI applications to automate complex, knowledge-intensive workflows - including open-ended decision-making, workflow execution, knowledge retrieval, and data transformation between business systems. Join the upcoming event series by Kore.ai [13] to stay on top of this emerging trend. Register
CaPa can generate high-quality 4K textured mesh under only 30 seconds. It can produce 3D assets for commercial applications such as games, movies, and VR/AR.
CityDreamer4D is a framework for creating high-fidelity 4D simulations of urban environments, enabling advanced planning and analysis in smart city applications.
AIRCHITECT V2 is a learning-based design space exploration tool developed to navigate the challenges of optimizing hardware accelerator designs for deep neural networks.
Bud-E is a system developed by Laion that can assist with educational queries in a more empathic and personalized way.
Intuition Robotics has launched a new ElliQ solution to assist caregivers of older adults. Featuring an AI-powered companion robot and a smartphone app, the system provides remote care management and helps alleviate caregiver stress by enabling caregivers to track health changes and set care goals. This innovation aims to enhance independence for older adults while transforming caregiving into a more rewarding experience.