FREE AI & ChatGPT Masterclass to automate 50% of your workflow

More than 300 Million people use AI across the globe, but just the top 1% know the right ones for the right use-cases.

Join this free masterclass on AI tools that will teach you the 25 most useful AI tools on the internet – that too for $0 (they have 100 free seats only!)

Get it now for absolutely free! (for first 100 users only) 🎁

This masterclass will teach you how to:

• Build business strategies & solve problems like a pro

• Write content for emails, socials & more in minutes

• Build AI assistants & custom bots in minutes

• Research 10x faster, do more in less time & make your life easier

You’ll wish you knew about this FREE AI masterclass sooner 😉

Register & save your seat now! (valid for next 24 hours only!)

Today:

• Carnegie Joins OpenAI Board

• Hugging Face Acquires XetHub Platform

• ChatGPT Expands DALL-E Access

• Meta's VFusion3D Advances 3D Modeling

• SoundHound Acquires Amelia for $80M

• Google Slashes Gemini 1.5 Costs

• Alibaba Tops AI Math with Qwen2

OpenAI's MEMETIC warfare... GPT-4o LARGE.

Highly persuasive AI models are emerging, capable of influencing public opinion through "mimetic warfare," where ideas spread like viruses. Smaller, efficient AI models like GPT-4 Mini are being rapidly developed and released. These models compress complex ideas into easily digestible forms, such as memes, making them more effective at swaying opinions. 

Speculation suggests that these AI models are already being tested in the wild, potentially reshaping how ideas are communicated and influencing societal beliefs. This could mark the beginning of an era where AI plays a significant role in steering public thought and perception.

WATCH THE VIDEO ON YOUTUBE

OpenAI adds a Carnegie Mellon professor to its board of directors

OpenAI has added Zico Kolter, a Carnegie Mellon professor specializing in AI safety, to its board of directors. This appointment follows a series of departures from the company, including safety-focused executives like co-founder Ilya Sutskever. 

Kolter will serve on OpenAI's Safety and Security Committee, which has faced criticism for being composed mostly of insiders. Kolter’s expertise in AI safety and his experience with industry collaborations are seen as valuable assets to OpenAI as it navigates challenges in governing advanced AI systems.

TECHCRUNCH

Hugging Face acquires XetHub from ex-Apple researchers for large AI model hosting

Hugging Face has acquired XetHub, a platform founded by ex-Apple researchers, to enhance its ability to host large AI models and datasets. This is Hugging Face’s largest acquisition to date. XetHub’s technology will be integrated into the Hugging Face platform, upgrading its storage capabilities and enabling support for much larger files than before. 

XetHub’s features, like content-defined chunking and deduplication, will streamline data handling, making it easier for teams to collaborate and manage large-scale AI projects. The acquisition aims to prepare Hugging Face for the future growth of AI models and datasets.

VENTUREBEAT

ChatGPT now lets free users generate up to two images per day made by DALL-E 3

OpenAI has announced that free users of ChatGPT can now generate up to two images daily using the DALL-E 3 model, a feature previously limited to ChatGPT Plus subscribers. DALL-E 3, launched last year, allows users to create images more easily by letting ChatGPT generate detailed prompts for them. 

The image creation feature is gradually being rolled out, with some users already able to access it. This update is part of a series of recent announcements from OpenAI, including a safety assessment of its GPT-4o model and the addition of a new board member.

THE VERGE

Meta’s VFusion3D: A leap forward in AI-powered 3D content creation

Meta and the University of Oxford have developed VFusion3D, an AI model that generates high-quality 3D objects from single images or text descriptions. This advancement addresses the challenge of limited 3D training data by using pre-trained video AI models to create synthetic 3D data. 

VFusion3D has shown impressive results, with human evaluators preferring its 3D reconstructions over previous models 90% of the time. The technology promises to revolutionize industries like gaming, virtual reality, and digital design by making 3D content creation more accessible and scalable, though challenges with specific object types remain.

VENTUREBEAT

SoundHound acquires Amelia AI for $80M after it raised $189M+

SoundHound, an AI company specializing in voice interfaces, has acquired Amelia AI, a customizable AI agent provider, for $80 million in cash and equity. Amelia AI had previously raised over $189 million. This acquisition allows SoundHound to expand into new sectors, including financial services and healthcare, and is expected to generate $150 million in revenue by 2025. The combined company will have 200 customers and $160 million in cash. 

Despite the lower acquisition price compared to Amelia’s funding, SoundHound sees significant potential in integrating Amelia’s technology to enhance its service offerings.

TECHCRUNCH

Gemini 1.5 Flash price drop with tuning rollout complete, and more

Google has announced significant updates for its Gemini 1.5 Flash model and API. The cost for using Gemini 1.5 Flash has been reduced, with input token prices dropping by 78% and output token prices by 71%. The model now supports over 100 languages, making it more accessible globally. Google AI Studio is now available by default for Google Workspace users, simplifying access. 

Additionally, developers can now fine-tune Gemini 1.5 Flash, improving performance and reducing costs. The Gemini API and AI Studio have also added PDF understanding capabilities, enhancing the processing of both text and visual content in PDFs.

GOOGLE FOR DEVELOPERS

Alibaba claims no. 1 spot in AI math models with Qwen2-Math

Alibaba Cloud has launched Qwen2-Math, a new series of math-specific AI models that now lead in global math performance benchmarks. The flagship model, Qwen2-Math-72B-Instruct, achieved an 84% score on the MATH Benchmark and outperformed competitors like OpenAI's GPT-4o and Google's Math-Gemini. These models excel in solving complex mathematical problems, including grade school and collegiate-level math, making them valuable tools for STEM fields. 

Qwen2-Math is open-source with some commercial usage restrictions, allowing wide accessibility for startups and enterprises. The release strengthens Alibaba's position in the AI landscape, particularly in math-intensive applications.

VENTUREBEAT

🧠RESEARCH

MiniCPM-V: A GPT-4V Level MLLM on Your Phone

MiniCPM-V is a highly efficient Multimodal Large Language Model (MLLM) designed for mobile and end-side devices. It outperforms major models like GPT-4V and Claude 3, offering strong OCR capabilities, high-resolution image perception, low hallucination rates, multilingual support, and efficient deployment on phones. This model signifies a trend toward reducing model sizes while enhancing performance, making advanced AI more accessible for real-world applications.

Lumina-mGPT: Illuminate Flexible Photorealistic Text-to-Image Generation with Multimodal Generative Pretraining

Lumina-mGPT is a multimodal autoregressive model that excels in generating photorealistic images from text. It uses a pretrained decoder-only transformer for multimodal token sequences. With Flexible Progressive Supervised Finetuning (FP-SFT) and Omnipotent Supervised Finetuning (Omni-SFT), it achieves versatile tasks, including text-to-image generation, segmentation, depth estimation, and visual question answering.

RAG Foundry: A Framework for Enhancing LLMs for Retrieval Augmented Generation

RAG Foundry is an open-source framework designed to simplify the implementation of Retrieval-Augmented Generation (RAG) systems for large language models. It integrates data creation, training, inference, and evaluation into a unified workflow, facilitating rapid prototyping and experimentation. Demonstrated by augmenting Llama-3 and Phi-3 models, RAG Foundry consistently improves performance on knowledge-intensive datasets. The code is available on GitHub.

MeshAnything V2: Artist-Created Mesh Generation With Adjacent Mesh Tokenization

MeshAnything V2 is an advanced autoregressive transformer that generates high-quality, artist-created meshes (AM) for 3D asset production. It uses a novel Adjacent Mesh Tokenization (AMT) method, which reduces token sequence length by about half and improves structure. This results in enhanced efficiency and performance compared to previous methods. Extensive experiments confirm the superiority of AMT in AM generation. 

VidGen-1M: A Large-Scale Dataset for Text-to-video Generation

VidGen-1M is a new, high-quality dataset for training text-to-video models. It addresses issues in existing datasets, such as poor temporal consistency and low video quality. Using a coarse-to-fine curation strategy, VidGen-1M ensures superior videos and captions. Models trained on VidGen-1M outperform those trained on other datasets, improving text-to-video generation significantly.

🛠️TOP TOOLS

Zoom Docs - AI-first customizable design easily adapts to team and project needs, including documents, wikis, and tables, providing a single place to manage work 

Black Forest Labs -  Image generation tool with exceptional performance, precise prompt following, and diverse output

Wondercraft - Create ads, podcasts, audiobooks - any audio content, in any language - just by typing.

Meta Sam 2 - Track and manipulate objects in images and videos effortlessly

📲SOCIAL MEDIA