OpenChat beats 100% of ChatGPT-3.5
Plus: AI designed CPU in <5 hours. SAM-PT: Video object segmentation.
Hello Engineering Leaders and AI Enthusiasts!
Welcome to the 55th edition of The AI Edge newsletter. This edition brings you “OpenChat beats 100% of ChatGPT-3.5”.
A big thank you to our incredible readers. We truly appreciate you! 😊
In today’s edition:
🤖 OpenChat beats 100% of ChatGPT-3.5
⚡️ AI designed CPU in <5 hours
🎯 SAM-PT: Video object segmentation with zero-shot tracking
📚 Knowledge Nugget: Improve Neural Networks by using Complex Numbers by
Let’s go!
OpenChat beats 100% of ChatGPT-3.5
OpenChat is a collection of open-source language models specifically trained on a diverse and high-quality dataset of multi-round conversations. These models have undergone fine-tuning using approximately ~6K GPT-4 conversations filtered from the ~90K ShareGPT conversations. It is designed to achieve high performance with limited data.
The model comes in three versions: The basic OpenChat model, OpenChat-8192 and OpenCoderPlus.
Why does this matter?
Despite the small amount of data, OpenChat is designed to achieve high performance. It demonstrates the potential for effective training with smaller datasets, advancing conversational AI and its applications in various fields like customer service and virtual assistants.
AI designs CPU in <5 hours
A team of Chinese researchers published a paper describing how they used AI to design a fully functional CPU based on the RISC-V architecture, which is as fast as an Intel i486SX. They called it a “foundational step towards building self-evolving machines.” The AI model completed the design cycle in under 5 hours, reducing it by 1000 times.
Why does this matter?
It's a major leap in self-evolving AI, as it sets the stage for AI systems to continually learn, adapt, and enhance their own designs. With the ability to significantly reduce the design cycle to 1000 times and optimize fine-grain architecture, AI-driven self-evolution becomes a tangible possibility.
SAM-PT: Video object segmentation with zero-shot tracking
Researchers introduced SAM-PT, an advanced method that expands the capabilities of the Segment Anything Model (SAM) to track and segment objects in dynamic videos. SAM-PT utilizes interactive prompts, such as points, to generate masks and achieves exceptional zero-shot performance in popular video object segmentation benchmarks, including DAVIS, YouTube-VOS, and MOSE. It takes a unique approach by leveraging robust and sparse point selection and propagation techniques.
To enhance the tracking accuracy, SAM-PT incorporates K-Medoids clustering for point initialization and a point re-initialization strategy.
Why does this matter?
SAM-PT’s ability to track and segment objects in dynamic videos with minimal prior information makes it a powerful tool in various applications, including video editing, surveillance, and autonomous systems.
Knowledge Nugget: Improve Neural Networks by using Complex Numbers
In this article, the author
discusses a concept called Hybrid Neural Networks and how they can make traditional Convolutional Neural Networks (CNNs) better! Also explores how using Complex Functions can take these models to the next level and improve their performance.Also provides an incredible and easily understandable explanation of the CoSh Network, delving into the comprehension and properties of hybrid neural networks and complex neural networks. Moreover, the article highlights the significance of why Complex Functions might be the next leap in Deep Learning.
Why does this matter?
Enhancing neural networks has the potential to boost the performance of AI models, while incorporating complex functions can contribute to the creation of versatile and resilient models applicable across diverse fields.
What Else Is Happening❗
📸 Midjourney introduces superb Panning feature which lets you explore Images in 360°! (Link)
🌍 AI and accelerated computing will help climate researchers achieve miracles! (Link)
🔠 Data scientists are using AI to translate Cuneiform & Akkadian into English. (Link)
🕺 Wow! DISCO can generate high-quality human dance images and videos. (Link)
👥 A real time full-body capture while being accurate in world space. (Link)
🛠️ Trending Tools
Wondercraft AI: AI-powered podcast builder, turns ideas into published podcasts within minutes.
Submagic: AI generates trendy captions with emojis for short-form content in under 2 minutes.
Goodmeetings: AI-curated reels condense meeting highlights, searchable library, human-like summaries.
FinanceGPT: SaaS platform combining generative AI, data-driven insights, visuals for financial decision-making.
Picker AI: AI photo picker app selects photos with high performance potential for various apps.
Ween AI: Turns customer data into actionable insights in minutes instead of weeks.
GPTsafe: 95% accurate text origin detection, AI rewrites text to sound human-authored.
Eazyeditor: AI-powered software transforms product images for photographers and e-commerce editors.
That's all for now!
Discover new opportunities and insights! By subscribing to The AI Edge, you become part of a select group of readers from Moody’s, Vonage, Voya, WEHI, Cox, INSEAD, and other reputable organizations.
Thanks for reading, and see you tomorrow. 😊