• TechTok Newsletter
  • Posts
  • Scarlett Johansson vs OpenAI 🎬, Meta's Multi-Modal Chameleon 🦎, Harvard's 3D Brain Reconstruction 🧠

Scarlett Johansson vs OpenAI 🎬, Meta's Multi-Modal Chameleon 🦎, Harvard's 3D Brain Reconstruction 🧠

Today's TechToks: Meta's paper on Chameleon, privacy in Slack's AI, issues with OpenAI's voice option, along with Trending GitHub repositories, the latest innovations in tech, Product Picks and more!

🤯 Game Changers

The most impactful articles of the day

Meta submitted a paper for their new family of Multi-Modal LLMs: Chameleon, capable of understanding and generating images and text in any arbitrary sequence. The paper shares a stable training approach, an alignment recipe, and an architectural parameterization, claiming Chameleon has state-of-the-art performance in image captioning tasks, performs non-trivial image generation, and is competitive with models such as Mixtral 8x7B, Gemini-Pro and GPT-4V.

Harvard and Google researchers published the largest dataset of neural connections, from 3D reconstructing of a cubic millimeter of human brain tissue with AI. The 'H01' dataset has over 57,000 cells, 230 millimeters of blood vessels, and 150 million synapses - amounting to 1,400 terabytes of data. The AI algorithms use Lichtman’s electron microscopy imaging to color-code and reconstruct the wiring of mammal brains, with the ultimate goal being to fully map a mouse’s neural wiring.

🧐 Daily Picks

Curated picks and most shared articles on techtok.today

Actress Scarlett Johansson refused OpenAI's offer to voice ChatGPT. But in the release date of latest flagship GTP-4o, the demo featured a similar voice and CEO Sam Altman tweeted 'her', a movie title starring Johansson— who hired legal counsel and spoke out, having previously won a lawsuit against Dinsney. OpenAI has since paused the use of the 'Sky' voice and apologized for not communicating better with Johansson.

Rethinking Text Resizing on Web

Airbnb rethinks text resizing on the web, prioritizing the use of rem units for font scaling to ensure a consistent and maintainable solution, while automating the unit conversion process to ease the transition for designers and developers across multiple CSS-in-JS systems.

✨ "Too long don't README"

GitHub trending repos of today

Bend is a massively parallel, high-level programming language that provides the features of expressive languages like Python and Haskell, while running on parallel hardware like GPUs with near-linear speedup based on core count and zero explicit parallel annotations.

The unofficial Oblivion secure and optimized internet access app for Desktop, using Cloudflare Warp technology, with a user-friendly interface and support for various platforms including Windows, macOS, and Linux.

Farfalle is an open-source AI-powered search engine that allows users to run their local large language models (LLMs) like llama3, gemma, and mistral, or use cloud models like Groq/Llama3 and OpenAI/gpt4-o, with features like Docker deployment, integration with LiteLLM, and support for SearXNG.

📱 Product Picks

Curated products from Product Hunt

Total control, infinite scalability, empower your team to swiftly adapt to changes and significantly reduce costs. Skip years of development and millions in investment just deploy NocoBase in minutes.

The best CMS is the one you already use. Connect to your existing CMS or the one that works best for you.

Your E-commerce Storyteller. Let your ideas come to life with captivating copy, engaging blogs, irresistible ads and much more—all powered by AI. Whether you're brainstorming for your next campaign or need fresh content, Smartli weaves the narrative for you.

Today from Engineering Blogs

Articles from engineering blogs of big tech companies

Slack built its AI features using large language models (LLMs) hosted on its own infrastructure, ensuring customer data never leaves Slack's control, and avoiding training models on customer data to protect privacy, while still leveraging the capabilities of advanced AI through a stateless Retrieval Augmented Generation approach.