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

Creating a universal definition of “open source AI” has generated a fair amount of debate and confusion, with many outfits using elastic parameters in order to achieve a fit. Now the Open Source Initiative (OSI) — “the authority that defines Open Source” — has issued what it hopes will become the baseline definition. That definition, which includes the ability to “use the system for any purpose and without having to ask for permission,” excludes a lot of AI platforms that currently describe themselves as “open,” many freely available only for non-commercial use. OSI’s remaining three parameters involve the ability to inspect the system and modify and share it. Continue reading OSI Aims for Industry Standard by Defining ‘Open Source AI’

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

Zoom is starting to launch its AI-powered tools first announced in October. Available to Zoom Workplace subscribers, the new Zoom Docs has been engineered from the ground up for AI optimization, leveraging Zoom’s AI Companion for what the company says will result in increased productivity and collaboration. Zoom Docs users will be able to open documents from within the videoconferencing app and can use generative AI to help write and edit them. The results will be easily shareable, Zoom says of its bid to compete with biggies like Google and Microsoft in the business productivity space. Continue reading Zoom Introduces AI-Powered Productivity Tools for Workplace

In April, Meta Platforms revealed that it was working on an open-source AI model that performed as well as proprietary models from top AI companies such as OpenAI and Anthropic. Now, Meta CEO Mark Zuckerberg says that model has arrived in the form of Llama 3.1 405B, “the first frontier-level open-source AI model.” The company is also releasing “new and improved” Llama 3.1 70B and 8B models. In addition to general cost and performance benefits, the fact that the Llama 3.1 405B model is open source “will make it the best choice for fine-tuning and distilling smaller models,” according to Meta. Continue reading Meta Calls New Llama the First Open-Source Frontier Model

A consortium of top tech firms have joined forces to launch a security group focused on artificial intelligence applications. The cybersecurity-focused non-profit OASIS will oversee operational aspects of the Coalition for Secure AI, to be known as CoSAI, described as an “open-source community.” OASIS lists Google, IBM, Intel, Microsoft, Nvidia and PayPal as founding Premier Sponsors of CoSAI, whose “additional founding sponsors” include Amazon, Anthropic, Cisco, Chainguard, Cohere, GenLab, OpenAI and Wiz. “CoSAI is an initiative to enhance trust and security in AI use and deployment,” OASIS announced at the Aspen Security Forum. Continue reading Google, OpenAI, Nvidia and Others Form AI Security Coalition

Amazon Web Services made availability announcements for services including its enterprise AI assistant Q, which now becomes available on its entry-level SageMaker tier, and introduced some new products at the AWS Summit at New York City’s Javits Center this week. Notably, the App Studio development assistant has launched in public preview. Amazon is also highlighting new features to improve AI accuracy, including a guardrail that detects “hallucinations.” Overall, the event — one in a series of daylong summits held in key cities across the nation — emphasized the comprehensiveness of the company’s generative AI stack. Continue reading AWS Expands Q Availability, Adds Guardrails for Bedrock AI

Cloudflare has a new tool that can block AI from scraping a website’s content for model training. The no-code feature is available even to customers on the free tier. “Declare your ‘AIndependence’” by blocking AI bots, scrapers and crawlers with a single click, the San Francisco-based company urged last week, simultaneously releasing a chart of frequent crawlers by “request volume” on websites using Cloudflare. The ByteDance-owned Bytespider was number one, presumably gathering training data for its large language models “including those that support its ChatGPT rival, Doubao,” Cloudflare says. Amazonbot, ClaudeBot and GPTBot rounded out the top four. Continue reading Cloudflare Blocking Web Bots from Scraping AI Training Data

San Francisco-based optics company Solos has debuted its latest smart glasses, the Solos AirGo Vision, which offer a camera that takes photos and provides computer vision, and integrates OpenAI’s GPT-4o. The AirGo Vision can provide real-time information using visual input, recognizing people, objects and places, and providing information such as directions or instructions. Both the camera and AI functionality are hands-free, making the AirGo Vision “especially convenient for visual progress and next steps on activities like cooking, home improvement projects, education and studies, and even shopping,” the company explains. Continue reading Solos AirGo Vision Smart Glasses Tout a Camera and GTP-4o

OpenAI has acquired Rockset, a database firm that provides real-time analytics, indexing and search capabilities. Rockset will help OpenAI enable its customers to better leverage their own data as they build and utilize intelligent applications. Rockset technology will be integrated into the retrieval infrastructure across OpenAI products, with members of Rockset’s San Mateo, California-based team joining the staff of OpenAI, which is headquartered in San Francisco. This is the second major purchase for OpenAI, following last year’s acquisition of New York-based AI design studio Global Illumination. Financial terms of the deal were not disclosed. Continue reading OpenAI to Expand Data Indexing, Analysis with Rockset Tech

MainFunc Inc. has raised $60 million on the strength of its principal technology — a free, AI-powered search engine called Genspark. The platform responds to queries by writing custom summaries that are presented in a “Sparkpage,” a one-page overview featuring content from around the web. Genspark joins a growing field of generative AI search engines, the best-known of which is Perplexity, which has raised $250 million since its 2022 launch and is currently valued at about $2.5 billion. Reuters says Genspark’s funding values the company at $260 million. Google also offers “AI Overviews” as part of Google search. Continue reading Genspark Joins Collection of GenAI-Powered Search Engines

Anthropic has launched a powerful new AI model, Claude 3.5 Sonnet, that can analyze text and images and generate text. That its release comes a mere three months after Anthropic debuted Claude 3 indicates just how quickly the field is developing. The Google-backed company says Claude 3.5 Sonnet has set “new industry benchmarks for graduate-level reasoning (GPQA), undergraduate-level knowledge (MMLU), and coding proficiency (HumanEval).” Sonnet is Anthropic’s mid-tier model, between Haiku and, on the high-end, Opus. Anthropic says 3.5 Sonnet is twice as fast as 3 Opus, offering “frontier intelligence at 2x the speed.” Continue reading Anthropic’s Claude 3.5: ‘Frontier Intelligence at 2x the Speed’

Ilya Sutskever — who last month exited his post as chief scientist at OpenAI after a highly publicized power struggle with CEO Sam Altman — has launched a new AI company, Safe Superintelligence Inc. Sutskever’s partners in the new venture are his former OpenAI colleague Daniel Levy and Daniel Gross, who founded the AI startup Cue, which was acquired by Apple where Gross continued in an AI leadership role. “Building safe superintelligence (SSI) is the most important technical problem of our​​ time,” the trio posted on the company’s one-page website, stating its goal is to “scale in peace.” Continue reading Sutskever Targets Safe Superintelligence with New Company

Nvidia is expanding its substantive influence in the AI sphere with Nemotron-4 340B, a family of open models designed to generate synthetic LLM training data for commercial applications across numerous fields. Through what Nvidia is calling a “uniquely permissive” free open model license, Nemotron-4 340B provides a scalable way for developers to build LLMs. Synthetic data is artificially generated data designed to mimic the characteristics and structure of data found in the real world. The offering is being called “groundbreaking” and an important step toward the democratization of artificial intelligence. Continue reading Nvidia’s Open Models to Provide Free Training Data for LLMs