The AI Canvas Newsletter #15
Explore the latest in AI news: Anthropic's Claude 3, Cognition's AI engineer Devin, Pi-2.5, Adobe's Music GenAI, and NVIDIA's StarCoder2
The AI Canvas Newsletter #15
The AI Canvas: Your weekly palette of inspiration, insights, and innovation in the world of AI.
- 🚀 Explore Claude 3: Anthropic's AI leap with Haiku, Sonnet, and Opus models, enhancing language and vision processing for diverse applications.
- 👨💻 Meet Devin: Cognition's autonomous AI software engineer, excelling in coding tasks and revolutionizing software development collaboration.
- 🧠Dive into Inflection-2.5: Inflection's personal AI update, rivaling top models with enhanced cognitive skills and empathetic user interactions.
- 🎶 Create with Adobe's Project Music GenAI Control: A groundbreaking tool for customizing music creation with text-driven editing capabilities.
- đź’» Utilize StarCoder2: ServiceNow, Hugging Face, and NVIDIA's open-access AI models, boosting developer productivity across 619 programming languages.
Written by Oli Wilkins.
Claude 3: The New Frontier in AI Model Performance and Application
The Claude 3 model family introduces three advanced AI models – Haiku, Sonnet, and Opus – each designed to cater to different levels of cognitive tasks and user needs. With enhanced capabilities in language comprehension, vision processing, and reduced refusal rates, these models promise improved performance for a variety of applications, from customer service to complex data analysis. The article details the models' features, benchmarks against competitors, and their commitment to responsible AI development.
Have a read on Anthropic’s blog and try the models out for yourself.
Devin: Pioneering AI in Software Engineering
Devin emerges as the first fully autonomous AI software engineer, demonstrating proficiency in complex engineering tasks and bug resolution. With a significant leap in performance on the SWE-bench coding benchmark, Devin offers real-time collaboration and autonomous problem-solving, allowing human engineers to tackle more complex challenges. Cognition, the team behind Devin, aims to expand AI's role in various disciplines, starting with coding.
Checkout the announcement here.
Introducing Inflection-2.5: Enhancing Personal AI with Superior IQ
Inflection-2.5, the latest update to the Pi personal AI, boasts significant advancements in cognitive capabilities, rivalling top language models like GPT-4 with only 40% of the computational power used for training. This update enhances Pi's empathetic interactions with users, offering improved performance in coding, mathematics, and real-time information retrieval, while maintaining its unique personality and safety standards. Users can now engage with Pi on a broader spectrum of topics, benefiting from its upgraded intelligence and efficiency.
Find out more here.
Adobe Unveils Project Music GenAI Control for Tailored Audio Creation
Adobe Research introduces Project Music GenAI Control, a tool enabling users to generate and finely edit music using text prompts. This innovation allows for detailed adjustments to tempo, structure, and intensity, streamlining the audio creation process for various content creators.
Read more here.
Introducing StarCoder2: Open-Access AI Models for Developer Productivity
ServiceNow, Hugging Face, and NVIDIA have released StarCoder2, a set of open-access large language models to support code generation in enterprise application development. With training across 619 programming languages, these models are available in different sizes to suit various computational needs and are designed with a focus on transparency and responsible AI usage. They aim to assist developers in improving efficiency and fostering innovation within their applications.
Read more here.
Technical Reads
Building Meta’s GenAI Infrastructure – Meta
“To lead in developing AI means leading investments in hardware infrastructure. Hardware infrastructure plays an important role in AI’s future. Today, we’re sharing details on two versions of our 24,576-GPU data center scale cluster at Meta.”
Beyond Self-Attention: How a Small Language Model Predicts the Next Token - Shyam Pather
“I trained a small (~10 million parameter) transformer following Andrej Karpathy’s excellent tutorial, Let’s build GPT: from scratch, in code, spelled out. After getting it working, I wanted to understand, as deeply as possible, what it was doing internally and how it produced its results.”
AI Watermarking 101: Tools and Techniques – Hugging Face
“In this blog post, we will describe approaches to carry out watermarking of AI-generated content, discuss their pros and cons, and present some of the tools available on the Hugging Face Hub for adding/detecting watermarks.”
Don't Mock Machine Learning Models In Unit Tests – Eugene Yan
“I’ve been applying typical unit testing practices to machine learning code and it hasn’t been straightforward. In software, units are small, isolated pieces of logic that we can test independently and quickly. In machine learning, models are blobs of logic learned from data, and machine learning code is the logic to learn and use these derived blobs of logic. This difference makes it necessary to rethink how we unit test machine learning code.
Training great LLMs entirely from ground up in the wilderness as a startup – Yi Tay
Given that we’ve successfully trained pretty strong multimodal language models at Reka, many people have been particularly curious about the experiences of building infrastructure and training large language & multimodal models from scratch from a completely clean slate.
The State of Competitive Machine Learning – ML Contests
“We summarise the state of the competitive landscape and analyse the 300+ competitions that took place in 2023. Plus a deep dive analysis of 60+ winning solutions to figure out the best strategies to win at competitive ML.”
Learning
“Welcome to “Feature Engineering A-Z”! This book is written to be used as a reference guide to nearly all feature engineering methods you will encounter. This is reflected in the chapter structure. Any question a practitioner is having should be answered by looking at the index and finding the right chapter.”
“The book introduces ideas from classical structural equation models (SEMs) and their modern AI equivalent, directed acyclical graphs (DAGs) and structural causal models (SCMs), and presents Double/Debiased Machine Learning methods to do inference in such models using modern predictive tools..”
🚀 Don't miss your weekly dose of cutting-edge AI innovations with The AI Canvas newsletter!
Subscribe now to ensure you never miss out on these transformative insights.
Looking for more specialised consultancy? At ADSP we’re a team of data experts who build AI products with purpose.
We deliver data science projects for companies who want to harness the power that AI can bring to their organisation. Get in touch at hello@adsp.ai.
Stay tuned with The AI Canvas podcast for in-depth episodes exploring Generative AI's transformative role across various industries.