- unwind ai
- Posts
- Last Week in AI - A Weekly Unwind
Last Week in AI - A Weekly Unwind
From 28-July-2024 to 3-Aug-2024
Shubham Saboo & Gargi Gupta
August 04, 2024
It was yet another thrilling week in the AI field with advancements that further extend the limits of what can be achieved with AI.
Here are 10 AI breakthroughs that you can’t afford to miss 🧵👇
Made using Whimsical Diagrams
Microsoft’s paper on using Knowledge Graphs for RAG highlights the high cost of constructing these graphs, as it requires generating an output token for every input token. SciPhi released Triplex, a new model that constructs knowledge graphs at 1/10th the cost of GPT-4o by extracting semantic triples directly from text. Triplex, available on HuggingFace and Ollama, outperforms GPT-4o in accuracy and integrates seamlessly with the R2R RAG engine and Neo4J.
Stanford Engineering and Toyota Research Institute have created a cutting-edge AI system capable of autonomously drifting two cars in tandem, with the trailing car reacting in real-time to the lead car's movements to prevent collisions. The AI utilizes a neural network vehicle model, learning from data to improve its performance over time, similar to a professional driver. This breakthrough aims to advance AI’s potential to improve safety in automated driving on public roads.
Meta has launched AI Studio which lets anyone create, share and discover AI characters. Creators on Instagram can build these AI characters as an extension of themselves to reach more fans. Built on Llama 3.1 models, you don’t need to have any technical skills to build these AI characters. You can pick a template to customize and make your own, or start entirely from scratch and build your AI from the ground up.
Apple has started rolling out Apple Intelligence via the new iOS 18.1 and macOS 15.1 in developer betas. These new OS will only be available for Macs and iPads with M1 chip or later, and iPhone 15 Pro and Max. Older devices will remain on iOS 18.0 beta or macOS 15.0 beta for now.
Meta has released SAM 2, the next generation of its Segment Anything Model for real-time object detection in both videos and images. Its predecessor SAM focused solely on image segmentation. SAM 2 can segment any object in any video or image, even if it hasn’t seen it before. It opens up a wide range of applications like video editing, mixed reality experiences, and various scientific fields.
OpenAI has started rolling out the new Voice Mode to a small group of ChatGPT Plus users. They will keep giving access to more people on a rolling basis. The video and screen-sharing capabilities of the Voice Mode won’t be available for now.
Google has expanded its Gemma 2 family with a new 2B parameter model and new tools to promote responsible AI development. The Gemma 2 2B model boasts impressive conversational AI capabilities, surpassing even GPT-3.5 models on the Chatbot Arena leaderboard. Alongside the new model, Google introduces ShieldGemma, a suite of safety classifiers to identify and mitigate harmful content, and Gemma Scope to get insights into the decision-making processes of Gemma 2.
Runway has trained a new video AI model called GEN-3 Alpha Turbo that can generate videos 7x faster than the original GEN-3 Alpha while matching its performance on many use cases. They will roll out Turbo for image-to-video with significantly lower pricing over the coming days and also make it available to free users.
GitHub has just launched GitHub Models in limited public beta, a built-in playground that lets you test leading AI models for free right within GitHub. From Llama 3.1 to GPT-4o and GPT-4o mini, to Phi 3 or Mistral Large 2, you can access each model in this playground and test different prompts and model parameters, before moving to a familiar coding environment in Codespaces and VS Code, and finally deploying with Azure AI for production-level support.
After Stable Diffusion, here’s a new strong contender in the opensource arena for text-to-image models. Black Forest Labs has just launched, along with their opensource image generation suite of models called FLUX.1. This suite of models delivers state-of-the-art performance in image detail, prompt adherence, style diversity, and scene complexity, outperforming even DALL.E-3 and Midjourney v6.
Which of the above AI development you are most excited about and why?
Tell us in the comments below ⬇️
That’s all for today 👋
Stay tuned for another week of innovation and discovery as AI continues to evolve at a staggering pace. Don’t miss out on the developments – join us next week for more insights into the AI revolution!
Click on the subscribe button and be part of the future, today!
📣 Spread the Word: Think your friends and colleagues should be in the know? Click the ‘Share’ button and let them join this exciting adventure into the world of AI. Sharing knowledge is the first step towards innovation!
🔗 Stay Connected: Follow us for updates, sneak peeks, and more. Your journey into the future of AI starts here!
Reply