Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Retool's State of AI report, which surveyed 1,500 tech workers, reveals generative AI's growing impact on business. MongoDB Atlas Vector Search leads in popularity among vector databases, vital for AI-powered apps enhancing productivity and customer experiences. However, challenges in AI adoption persist, notably in model accuracy and data security, with C-suite executives showing greater optimism than individual contributors.
You.com has launched APIs starting at $100 per month, enabling large language models like Meta’s Llama 2 to access up-to-date internet data and offering enhanced answer accuracy through web search and news functionalities.
At OpenAI’s DevDay, Microsoft announced plans to offer the new GPT-4 Turbo on Azure OpenAI Service by year-end, providing enhanced control and efficiencies at lower costs. OpenAI's Custom Models program will seamlessly integrate with Microsoft's ecosystem, boosting enterprises’ AI capabilities.
Nous research has released a very strong 34B model built on top of the bilingual Yi model. Capybara is trained primarily on multi-turn data with more than 1000 tokens per conversation, which makes it uniquely positioned for certain user-facing tasks with high information density.
When new models come out with astounding capabilities, it is natural to question whether they actually trained on the validation data or on benchmarks. Decontamination is the process of removing such testing data from the input. The LMSYS group found that if you rephrase test data in a way that passes decontamination but still contains information about the benchmark, you can fake your way to amazing performance.
This study introduces MonoDiffusion, a novel framework for self-supervised monocular depth estimation, which uniquely approaches the challenge as an iterative denoising process. It utilizes a pseudo ground-truth diffusion process guided by a pre-trained teacher model to improve depth map accuracy without using actual depth ground-truth in training.
Sponsored
Need to deliver real-time, low-latency data services to edge-based applications? Use NATS.io [13] — the same technology [13] used by thousands of companies across leading Financial Services, Automotive, Retail, and Government institutions. ➡️ Process massive amounts of data [13] where it is generated in real-time
➡️ Focus [14] on the AI application, not the cloud infrastructure
A compelling demo that uses GPT-V and a pleasant sketching interface to create HTML websites.
This project introduces Knowledgeable Preference AlignmenT (KnowPAT), a novel approach to improving domain-specific question answering using large language models.
This project introduces AMBER, a new benchmark designed to assess and reduce hallucinations in Multi-modal Large Language Models (MLLMs) without relying on other advanced language models.
IBM has committed $500 million to invest in generative AI startups that cater to business clients, signaling a strategic shift and indicating its readiness to support emerging companies, with a preference for those that complement rather than compete with its own business offerings.
Humane's AI Pin is a $699 screenless wearable device with innovative interaction methods powered by a $24 monthly subscription that offers features like language translation, AI-assisted photography, and personal assistant tools on T-Mobile's network.
A study by Redwood Research reveals that large language models (LLMs) can master "encoded reasoning," a form of steganography allowing LLMs to embed intermediate reasoning steps in generated text, making them undecipherable to humans.
Samsung has unveiled Samsung Gauss, a generative AI model that integrates with its devices and enhances tasks like email composition, software development, and image processing.
An AI applicant tracking and hiring tool.
Structured extraction in Python powered by OpenAI’s function calling API.
Sponsored
Anyscale [5] is the company behind the open-source framework called Ray, used by the world’s leading AI teams.
Nvidia is launching the HGX H200, an advanced AI chip with significantly more memory bandwidth and capacity than its predecessor, the H100. Availability remains uncertain as Nvidia works with partners for its 2024 release.
OpenAI is working on GPT-5 and seeking additional funding from Microsoft Corp. for its development.
Xbox and Inworld AI are collaborating to create AI-powered tools to help enrich the narrative and character creation elements in game development. The partnership will develop an AI design copilot to assist game designers and an AI character runtime engine for an immersive gaming experience. They anticipate these tools will foster limitless creativity, enhance immersion, and expedite game production.
Advances in AI, including AlphaFold, have significantly increased the rate of progress in science. This study presents RoseTTAFold All-Atom, a neural network that can model a wide range of molecules beyond proteins including drugs, vitamins, DNA, and sugars. The authors validate their architecture by fine-tuning it into a diffusion model and experimentally validating new-to-nature molecules designed from minimal prompts across many applications.
This paper introduces FinGPT, a language model trained on Finnish. The researchers who created it found that data repetition with some tricks produced extremely smooth loss curves. This may serve as an easy fix for underrepresented languages on the internet.
VLPMarker, a new watermarking method, protects Embedding as a Service Vision-Language Pre-Trained Models from extraction attacks.
Sponsored
Mendable provides AI chat search [13] for any use case — documentation, support, product copilots, and more. It’s as fast and simple as _npm install @mendable/search_ [13]. ✔️ Trusted by Snap, MongoDB, Worldline
An open-source project building upon LangChain, LangServe, and LangSmith that offers customizable language models, prompts, tools, vector databases, retrieval algorithms, and chat history databases. It features a sandbox environment, custom actions, and analytics for enhanced user control similar to OpenAI's GPTs.
The enhanced LLaMA model now includes 16,000 Tamil tokens, leveraging the LoRA methodology for efficient training. This update, along with new Tamil versions of Alpaca and OpenOrca datasets, has notably improved Tamil text processing, paving the way for future advancements in Indian language AI.
This report evaluates GPT-4V for autonomous driving, focusing on its capabilities in scene understanding, decision-making, and real-time responses.
Sparse tensors occur when some values of the neural network are zeros. These values can efficiently be skipped. Researchers developed a way to quickly find when values are zero and overload hardware to fit more non-zero values onto a chip at a time.
A comprehensive review of the major developments in the AI research, industry, and open-source space that happened in 2023.
Sponsored
This notebook combines the tactile satisfaction of pen-on-paper with the sheer convenience and power of AI. Become a backer to get the
OpenAI's six-member board will decide when AGI is achieved, potentially impacting its commercial relationship with Microsoft.
Civitai, an AI model sharing platform, has introduced a controversial "bounties" feature, allowing users to request and monetize the creation of AI-generated images.
Top-ranked OpenAI GPTs.
Google is in talks to deepen its relationships with Character.AI by investing hundreds of millions of dollars.
Zapier introduced AI Actions, a tool for developers to let any AI platform run Zapier's 20,000+ automation actions. AI Actions works by letting users send natural language commands to the AI platform, which then performs the desired action. The service supports several AI platforms, with simple setup and inherent API integrations.
Figma has incorporated AI assistance into FigJam, its digital whiteboard tool, to simplify and enhance design collaborations. Utility-oriented enhancements, like those derived from the AI-powered project Jambot, help users collaborate more effectively on a virtual canvas. Figma’s goal is to broaden the applicability across various user requirements by tapping machine learning capabilities for visual design.
This blog post is about different attacks that are emerging against language model systems. It contains excellent information about different types of attacks and some mitigations that teams have found to be effective.
This research presents the 3DStyle-Diffusion model, a novel approach for detailed stylization of 3D meshes, integrating 2D Diffusion models for added control over appearance and geometry. It works by first parameterizing a 3D mesh's texture into reflectance and lighting, using implicit MLP networks, and then using a pre-trained 2D Diffusion model to align the rendered images with the text prompt and ensure geometric consistency.
Sponsored
Rigorous unit testing for ML models [11] increases trust in your AI/ML systems, and enables deeper insights into model performance in a fraction of the time. Discover how it’s done in this upcoming webinar by Kolena [11]. You’ll learn how to design a model quality management process; standardize test coverage; and deal with robustness, fairness, and
With the recent release of the excellent Zephyr language model, the HuggingFace team showcases how you can train personalized models built on top of the few powerful pre-trained open source models available.
Audio super-resolution is the process of increasing the quality and fidelity of any audio, real or synthetic. Most super-resolution systems are task-specific, with single models trained for single audio data types (e.g., speech vs music). This new work is an amazing step forward where a single model can serve to increase the quality of audio across tasks.
With the advent of powerful new vision models, many groups are attempting to build agents that use vision to interact with web elements. Tarsier toolkit introduces a standard set of tools (e.g., element tagging). You can use any vision system to understand the web page and take action. It also includes utilities for non-vision language models to browse.
An example of using Bark text-to-speech system to generate coherent, fast, and extra-long audio output.