Nvidia has debuted a new AI model, Llama-3.1-Nemotron-70B-Instruct, that it claims is outperforming competitors GPT-4o from OpenAI and Anthropic’s Claude 3.5 Sonnet. The impressive showing has prompted speculation of an AI shakeup and a significant shift in Nividia’s AI strategy, which has thus far been focused primarily on chipmaking. The model was quietly released on Hugging Face, and Nvidia says as of October 1 it ranked first on three top automatic alignment benchmarks, “edging out strong frontier models” and vaulting Nvidia to the forefront of the LLM field in areas like comprehension, context and generation. Continue reading Nvidia’s Impressive AI Model Could Compete with Top Brands

Generative video models seem to be debuting daily. Pyramid Flow, among the latest, aims for realism, producing dynamic video sequences that have temporal consistency and rich detail while being open source and free. The model can create clips of up to 10 seconds using both text and image prompts. It offers a cinematic look, supporting 1280×768 pixel resolution clips at 24 fps. Developed by a consortium of researchers from Peking University, Beijing University and Kuaishou Technology, Pyramid Flow harnesses a new technique that starts with low-resolution video, outputting at full-res only at the end of the process. Continue reading Pyramid Flow Introduces a New Approach to Generative Video

Nvidia has unveiled the NVLM 1.0 family of multimodal LLMs, a powerful open-source AI that the company says performs comparably to proprietary systems from OpenAI and Google. Led by NVLM-D-72B, with 72 billion parameters, Nvidia’s new entry in the AI race achieved what the company describes as “state-of-the-art results on vision-language tasks, rivaling the leading proprietary models (e.g., GPT-4o) and open-access models.” Nvidia has made the model weights publicly available and says it will also be releasing the training code, a break from the closed approach of OpenAI, Anthropic and Google. Continue reading Nvidia Releases Open-Source Frontier-Class Multimodal LLMs

AI startup Liquid, founded by alums of MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), has released its first models. Called Liquid Foundation Models, or LFMs, the multimodal family approaches “intelligence” differently than the pre-trained transformer models that dominate the field. Instead, the LFMs take a path of “first principles,” which MIT describes as “the same way engineers build engines, cars, and airplanes,” explaining that the models are large neural networks with computational units “steeped in theories of dynamic systems, signal processing and numeric linear algebra.” Continue reading MIT Spinoff Liquid Eschews GPTs for Its Fluid Approach to AI

The Allen Institute for AI (also known as Ai2, founded by Paul Allen and led by Ali Farhadi) has launched Molmo, a family of four open-source multimodal models. While advanced models “can perceive the world and communicate with us, Molmo goes beyond that to enable one to act in their worlds, unlocking a whole new generation of capabilities, everything from sophisticated web agents to robotics,” according to Ai2. On some third-party benchmark tests, Molmo’s 72 billion parameter model outperforms other open AI offerings and “performs favorably” against proprietary rivals like OpenAI’s GPT-4o, Google’s Gemini 1.5 and Anthropic’s Claude 3.5 Sonnet, Ai2 says. Continue reading Allen Institute Announces Vision-Optimized Molmo AI Models

Artificial intelligence platform Runway has launched The Hundred Film Fund to help finance 100 projects that use its AI to tell stories. Created by the company through its Runway Studios, the Fund is starting with $5 million, “with the potential to grow to $10 million.” Runway is presenting the Fund as “an open call to all creative professionals who have AI-augmented film projects in the pre- or post-production phases and are in need of funding.” Directors, producers and screenwriters are among those invited to apply. The program will consider all formats, from features to shorts, documentaries, experimental projects, music videos and more. Continue reading Runway Launches $5M AI Film Fund as Open Call to Creators

Cloudflare has released AI Audit, a free set of new tools designed to help websites analyze and control how their content is used by artificial intelligence models. Described as “one-click blocking” to prevent unauthorized AI scraping, Cloudflare says it will also make it easier to identify the content bots scan most, so they can wall it off and negotiate payment in exchange for access. Helping its clients toward a sustainable future, Cloudflare is also creating a marketplace for sites to negotiate fees based on AI audits that trace cyber footprints on server files. Continue reading Cloudflare Tool Can Prevent AI Bots from Scraping Websites

Amazon is transferring its OpenSearch platform to the Linux Foundation’s new OpenSearch Software Foundation. By handing a third-party the open-source project it has developed internally since 2021, Amazon hopes to accelerate collaboration in data-driven search and analytics, an area of focus due to the proliferation of model training. Not to be confused with commercial search (Google, Bing), engines like OpenSearch are geared toward enterprise and academia. Because it is licensed under Apache 2.0, OpenSearch is a viable starting point for organizations that customize internal platforms for searching, monitoring and analyzing large volumes of data. Continue reading AWS Transfers OpenSearch Stewardship to Linux Foundation

OpenAI is previewing a new series of AI models that can reason and correct complex coding mistakes, providing a more efficient solution for developers. Powered by OpenAI o1, the new models are “designed to spend more time thinking before they respond, much like a person would,” and as a result can “solve harder problems than previous models in science, coding, and math,” OpenAI claims, noting that “through training, they learn to refine their thinking process, try different strategies, and recognize their mistakes.” The first model in the series is being released in preview in OpenAI’s popular ChatGPT and in the company’s API. Continue reading OpenAI Previews New LLMs Capable of Complex Reasoning

YouTube is introducing AI detection tools designed to allow people to learn when their face and/or voice are copied and used in third-party videos. As part of the effort, YouTube’s existing Content ID program that protects copyrighted music will expand to include more broad-based voice simulation detection technology. The new tools aim to protect “people from a variety of industries — from creators and actors to musicians and athletes,” according to the company. The Google-owned platform is also coming up with a way to address unauthorized use of its content for training AI models. Continue reading YouTube Adding Tools to Protect Against Unauthorized AI Use

In a move toward increased transparency, San Francisco-based AI startup Anthropic has published the system prompts for three of its most recent large language models: Claude 3 Opus, Claude 3.5 Sonnet and Claude 3 Haiku. The information is now available on the web and in the Claude iOS and Android apps. The prompts are instruction sets that reveal what the models can and cannot do. Anthropic says it will regularly update the information, emphasizing that evolving system prompts do not affect the API. Examples of Claude’s prompts include “Claude cannot open URLs, links, or videos” and, when dealing with images, “avoid identifying or naming any humans.” Continue reading Anthropic Publishes Claude Prompts, Sharing How AI ‘Thinks’

OpenAI announced its newest model, GPT-4o, can now be customized. The company said that the ability to fine-tune the multimodal GPT-4o has been “one of the most requested features from developers.” Customization can move the model toward more specific structure and tone of responses or allow it to follow specific instruction sets geared toward individual use cases. Developers can now implement custom datasets, aiming for better performance at a lower cost. The ChatGPT maker is rolling out the welcome mat by offering 1 million training tokens per day “for free for every organization” through September 23. Continue reading OpenAI Pushes GPT-4o Customization with Free Token Offer

Meta Platforms CEO Mark Zuckerberg and Spotify CEO Daniel Ek have joined forces to express displeasure with the European Union’s regulations on artificial intelligence, claiming they are suppressing innovation. That is the opposite of the stated goals of EU lawmakers in passing the regulations. In a joint statement first published in The Economist and then on the Meta and Spotify websites Friday, the duo took aim at alleged EU obstruction to the development of open source AI, suggesting that Europe’s “fragmented regulatory structure, riddled with inconsistent implementation, is hampering innovation and holding back developers.” Continue reading Meta, Spotify Issue Statement Criticizing EU’s AI Regulations

Palo Alto-based startup PIP Labs announced an $80 million funding round for Story Protocol, a blockchain platform to track intellectual property rights in the era of artificial intelligence and the data scraping that enables model training. CEO and co-founder Seung Yoon “SY” Lee says the company aims to create a more sustainable IP environment for digital consumers and builders. The raise, led by Andreessen Horowitz (a16z) and Polychain Capital, values the startup at $2.25 billion. The move comes after Sahara AI announced it raised $43 million this month to fund a blockchain-based IP tracking system. Continue reading Story Raises $80M to Create Blockchain-Based IP Protection

Google DeepMind has made its latest AI image generator, Imagen 3, free for use in the U.S. via the company’s ImageFX platform. Imagen 3 will be available in multiple versions, “each optimized for different types of tasks, from generating quick sketches to high-resolution images.” Google announced Imagen 3 at Google I/O in March, and in June made it available to enterprise users through Vertex. Using simplified natural language text input rather than “complex prompt engineering,” Google says Imagen 3 generates high-quality images in a range styles, from photorealistic, painterly and textured to whimsically cartoony. Continue reading Google DeepMind Releases Imagen 3 for Free to U.S. Users