Streamline your development process with Pinata’s easy File API

• Easy file uploads and retrieval in minutes

• No complex setup or infrastructure needed

• Focus on building, not configurations

Try today!

Today:

• OpenAI's o1 Model Trademarked

• Grok Chatbot Expands with App

• Bluesky API Sparks Data Concerns

• GenFM Turns Text to Podcasts

• Hume's EVI 2 Voice PC Commands

OpenAI's o1 Model Trademarked

OpenAI is trademarking its new AI model, o1, to protect its intellectual property. The company filed for a trademark with the U.S. Patent and Trademark Office (USPTO) and had previously applied for one in Jamaica before announcing o1. The USPTO has not yet approved the trademark. 

OpenAI says o1 is a “reasoning” model designed to self-check and avoid common AI errors. The company has faced trademark challenges before, notably failing to secure "GPT" as it was deemed too generic. OpenAI is also in a legal battle with Guy Ravine over the use of “Open AI.”

Grok Chatbot Expands with App

Elon Musk's AI company, xAI, plans to launch a standalone app for its Grok chatbot to compete with OpenAI's ChatGPT. This move marks xAI's next step in the AI market, as the company races to expand its capabilities. Currently, users can only access Grok through X with a subscription. 

xAI also supports customer service for Musk's other company, SpaceX. With this new app, xAI aims to join the ranks of AI chatbots that already have dedicated apps, such as OpenAI's ChatGPT, Google's Gemini, and Anthropic's Claude.

Bluesky API Sparks Data Concerns

Bluesky, a social platform, faces scrutiny over its open API, which allows third parties to scrape user data for AI training. Recently, Daniel van Strien from Hugging Face used Bluesky's API to collect 1 million public posts for research, sparking controversy. Though Bluesky does not train AI on user content, it acknowledges that public posts are accessible to anyone. 

The company is exploring ways to let users set consent preferences for data use, but enforcement relies on external developers. As Bluesky's popularity grows, it is under increasing scrutiny, similar to other major social networks.

GenFM Turns Text to Podcasts

ElevenLabs' GenFM on ElevenReader turns your content into engaging podcasts with AI co-hosts. Simply import your articles, documents, PDFs, eBooks, or newsletters to generate a customized podcast. Featuring realistic AI co-hosts tailored to your content, the service supports 32 languages.

Whether you're catching up on the latest news, diving into book reviews, preparing study notes, or turning downtime into story time, GenFM makes it accessible anywhere. You can download the iOS app now, with an Android version coming soon.

Hume's EVI 2 Voice PC Commands

Hume's EVI 2 technology allows natural voice control for computers. It processes spoken commands instantly and sends instructions to the computer, explaining actions verbally and allowing interruptions to change tasks. Built on Replit’s template and using AnthropicAI's API, EVI 2 can both generate its own speech and read lines from other language models.

It works with any language model and is available as an API. This innovation makes computer control more intuitive and user-friendly by enabling seamless voice interaction.

🧠RESEARCH

ShowUI: One Vision-Language-Action Model for GUI Visual Agent

ShowUI is a new model for digital assistants that better understands screen visuals, improving user experience. Key features include reducing redundant data, efficiently processing tasks, and using high-quality training data. ShowUI is lightweight but effective, achieving 75.1% accuracy in tasks and improving speed by 1.4 times.

Star Attention: Efficient LLM Inference over Long Sequences

Star Attention improves the efficiency of large language models (LLMs) by reducing computational costs and memory usage. It uses a two-phase approach: first, processing context locally in parallel across multiple hosts, and then attending to all prior tokens globally. This method speeds up inference by up to 11 times while maintaining high accuracy.

Material Anything: Generating Materials for Any 3D Object via Diffusion

Material Anything is an automated framework for generating high-quality materials for 3D objects under diverse lighting conditions. It uses a pre-trained image diffusion model with advanced features to ensure stability and material quality. The approach includes confidence masks and a UV-space material refiner, outperforming existing methods in various scenarios.

Large-Scale Text-to-Image Model with Inpainting is a Zero-Shot Subject-Driven Image Generator

Large-Scale Text-to-Image Model with Inpainting introduces Diptych Prompting, a zero-shot method for generating images of new subjects in specific contexts without extensive fine-tuning. It uses inpainting to align subjects accurately, improves detail through enhanced attention, and prevents unwanted content leakage. This method excels in subject-driven generation, stylized image creation, and editing.

GMAI-VL & GMAI-VL-5.5M: A Large Vision-Language Model and A Comprehensive Multimodal Dataset Towards General Medical AI

GMAI-VL & GMAI-VL-5.5M present a new model and dataset to enhance medical AI by integrating visual and textual information. This approach improves diagnosis and clinical decisions. The dataset, GMAI-VL-5.5M, combines image-text pairs from hundreds of medical datasets. The model excels in tasks like visual question answering and medical image diagnosis.

🛠️TOP TOOLS

Socap - AI networking copilot for entrepreneurs

Toolhouse - Complete cloud infrastructure to equip LLMs with actions and knowledge.

Lune AI - AI For CodeThat Actually Reads The Docs.

Video Ocean - Create videos from text and images

Hume - Combine voices and personalities generated by our speech-language model, EVI 2, with supplemental LLMs and tools like the new Claude 3.5 Haiku

🗞️MORE NEWS

• Elon Musk announced plans to start an AI game studio to "make games great again." He believes too many studios are owned by large corporations. Musk has shown a deep interest in gaming, including being a top 20 Diablo 4 player and a successful streamer.

• SoftBank is investing $1.5 billion in OpenAI through a tender offer, letting employees sell shares. OpenAI has raised significant funds and will allow more secondary share sales for employees. The deal is not tied to OpenAI's restructuring plans.

• Google's Gemini AI can now play music from Spotify using voice commands on Android. Users can search by song title, artist, album, playlist, or activity. The feature requires linking Spotify and Google accounts, enabling Gemini Apps Activity, and is initially available only in English.

• Google has launched a new chess website where AI customizes pieces based on user descriptions. The game, although not fully featured, offers a unique visual experience. This release coincides with the 2024 World Chess Championship and includes the announcement of a new chess bot within Google's AI, Gemini.

• AI agents in Minecraft developed complex social behaviors on their own, forming jobs, sharing memes, and spreading religion. Altera's experiment, involving up to 1000 agents using large language models, highlights the potential for AI civilizations. Founder Robert Yang envisions AI agents collaborating with humans in digital spaces.