Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Systems powered by Nvidia's Hopper architecture dominated the results of two new tests from MLPerf, an AI benchmarking suite that compares the fine-tuning of large language models and training of graph neural networks.
Perplexity is leveraging its strategic partnership with SoftBank to expand its reach in Japan. It is offering a free year of its premium AI-powered search engine, Perplexity Pro. SoftBank aims to attract users with AI services without developing in-house solutions. Perplexity, valued at $1 billion, is advancing its funding and investor base, which includes notable tech leaders and venture firms.
ERASE is a new approach for updating language models. Unlike traditional methods that focus on improving retrieval during prediction, ERASE incrementally deletes or rewrites entries in the knowledge base when new documents are added.
CAMixerSR performs image super-resolution by smartly applying convolution for simpler parts and deformable window-attention for detailed textures.
Sponsored
Want to keep your personal details out of the next data breach? Start by not giving them to every nosy website. Protect your real identity with Surfshark VPN + Alternative ID for <$2.79/mo [12].
Chameleon is an early fusion model that operates on images and text tokens simultaneously. Its team released the paper a few weeks ago. They have now released model checkpoints along with inference code.
OGNI-DC is a new framework for depth completion using "Optimization-Guided Neural Iterations" (OGNI). It refines a depth gradient field and integrates the depth gradients into a depth map.
Subobject tokenization is a new way for vision models to understand images. Instead of breaking images into fixed square patches, models using subobject tokenization look at images in terms of meaningful segments, like parts of objects.
Open source local agent Open Interpreter has just had a big upgrade. It can control the computer seamlessly and works completely offline and locally.
LlamaIndex introduces the Property Graph Index, enhancing knowledge graph capabilities with better modeling, storage, and querying features. The new index allows flexible construction of graphs, supports schema-guided, implicit, and free-form entity extraction, and integrates with vector databases for hybrid searches. It provides querying through keyword expansion, vector similarity, Cypher queries, and custom traversal.
Open models with state-of-the-art code performance that do depth upscaling to initialize larger models. Unlike many code models, these models also excel at fixing code, explaining code, and evaluating areas for improvement.
Sponsored
Bots and fraudulent sign-ups are wasting your time. Get back to innovation with Clerk's designed-for-AI login and permissions
Decagon is building human-like AI agents for customer support. It raised a $30M Series A from Accel and $5M Seed from a16z. Decagon's product handles global support for companies like Eventbrite, Rippling, Webflow, BILT, and Substack.
Open model that matches GPT4-o on function calling benchmarks trained on top of Llama 3 70B.
Experience increased productivity and accelerated growth with AI-powered features integrated into the HubSpot customer platform.
Runway has trained an extremely powerful new video generation model. It will power many of the existing features on its platform. Examples are available in the provided link.
DeepMind is developing an AI technology called V2A to generate synchronized soundtracks for videos. It uses diffusion models trained on audio, dialogue transcripts, and video clips to create music, sound effects, and dialogue.
Cerebras, a California-based company, has demonstrated that its second-generation wafer-scale engine is significantly faster than the world's faster supercomputer in molecular dynamics calculations. It can also perform sparse large language model inference at one-third of the energy cost of a full model without losing any accuracy. Both achievements are possible due to the interconnects and fast memory access enabled by Cerebras' hardware. Cerebras is looking to extend the applications of its wafer-scale engine to a larger class of problems, including molecular dynamics simulations of biological processes and simulations of airflow around vehicles.
The new Depth Anything model was trained primarily on synthetic data and has dramatically improved performance on complex scenes.
DeepSeek Coder is a powerful model that gets 90+ on HumanEval while matching GPT-4 Turbo performance on many other challenging benchmarks. It is available through an API and free for commercial use.
Nvidia has released a dataset and recipe along with a high quality paper about training reward models to align model output to human preferences.
Sponsored
Enhance your search experience with AI while maintaining flexibility and deep control over your architecture and UI. Read the blog [10] for a tech overview, or visit the Algolia Code Exchange [11] to access frontend and UI components, and to explore open source AI search
Everything should be differentiable. This post walks through how to write differentiable SVG lite.
LARS is an application that enables you to run LLMs locally on your device. Upload your own documents and engage in conversations where the LLM grounds its responses with your uploaded content.
This paper investigates why diffusion-based image generation models create "hallucinations" — images that never appeared in the training data.
The creator of RAGatouille gave an excellent talk discussing how to dramatically improve RAG performance, some of the open challenges, and COLBERT.
Chain of Preference Optimization (CPO) is a method that improves the logical reasoning abilities of large language models (LLMs). By fine-tuning LLMs using search trees from the Tree-of-Thought (ToT) method, CPO aligns the reasoning steps of Chain-of-Thought (CoT) decoding with ToT's optimal paths.
What would it take to make a generally intelligent agent and what are we missing? This post explores the 3 ideas needed to make an agent and posits that we are only a few years away. The author is a researcher at OpenAI.
TokenCost helps calculate the USD cost of using major LLM APIs by calculating the estimated cost of prompts and completions.
MindMatrix is an AI-powered mind map generator that simplifies the creation and visualization of complex concepts.
Install and run your own AI agent service.
NVIDIA's Nemotron-4 340B is a family of open models that developers can use to generate synthetic data for training LLMs for commercial applications. The state-of-the-art reward model matches the original GPT-4 model and can run on 8 H100s.
Apple and OpenAI's partnership isn't expected to generate meaningful revenue for either party, at least at the beginning. Apple believes that pushing OpenAI's brand and technology to its devices is of equal or greater value than monetary payments so it isn't paying OpenAI as part of the deal. The deal isn't exclusive - Apple is already discussing offering other chatbots as additional options. Apple eventually plans to make money from AI by striking revenue-sharing agreements with AI partners.
This article contains a transcript of an interview with Aidan Gomez, CEO of Cohere, where he discusses Cohere's path to profitability, the benefit of competition in the enterprise space, AI's current abilities, the future of large language models, and more.
Simple diffusion model for language modeling. It generates out of order and works quite well.
Researchers are addressing the vulnerability of Large Language Models to jailbreak attacks, which manipulate models to produce harmful outputs.
Multi-Layer Key-Value (MLKV) sharing is a new method that significantly reduces memory usage during auto-regressive inference in transformers.
The current Holy Grail of research is figuring out how to integrate search into language model generation. This work is extremely promising and shows that small models with search can match much more powerful models on math performance.
VideoGPT+ is a model that combines image and video encoders to enhance video understanding. Image encoders capture detailed spatial information, while video encoders provide temporal context.
This project introduces a new dataset and framework to improve Scene Graph Generation (SGG) for very-high-resolution satellite imagery (VHR SAI).
Aiyana Rainbow, a Romanian AI model, is a finalist in the inaugural Miss AI pageant, an event that features AI-generated models on social media. Organized by FanVue, the event is part of "The FanVue World AI Creator Awards," which showcases AI creators' ability to generate engaging content without being the physical face themselves. The winner of Miss AI is set to receive $5,000, public relations support, and mentorship. The results will be announced at the end of June.
Governments around the world are boosting their budgets to encourage companies and multinationals to build new data centers and refit old ones to develop AI locally and train large language models in their native languages based on their own citizens' data. Nvidia says that these sovereign AI efforts are expected to bring in almost $10 billion this year. Many governments are concerned about how generative AI will impact their economies. They want sovereign clouds for their AI infrastructure and sensitive data, and US tech companies are eager to build those for them.
Elon Musk responded to Apple's integration of OpenAI ChatGPT on some of its devices by threatening to ban all Apple devices from the premises of his companies.
OpenAI, initially created to counter DeepMind's AI advancements, is now closely partnered with Microsoft, which has rights to OpenAI's technology due to significant investments.
A port of Andrej Karpathy's llm.c to Mojo.
Amazon says that it will commit up to $230 million to startups building generative AI-powered applications.