📆 June 28th 2024📆 

New models and features from the big names in AI & the first use of GenVideo in advertising

• Google announces Gemma2 as it’s latest model in 9B and 27B variants

• OpenAI are automating quality checks on ChatGPT using CriticGPT

• Toys ‘R’ Us become the first company to release an AI Generated ad campaign using OpenAI’s Sora model.

• YouTube are negotiating with music artists to use their content in new AI ventures.

• As always, too much to put in headlines. Check out the Other News Rollup section for a quickfire review of the top stories in AI.

Dive in for all the details!

🆘 Quick Tips 🆘

Today I’m going to breakdown how to create the perfect image using any of the image generators. There may be slight variation between systems but use these approaches will help you craft the image you want. The overarching tip, imaging you are the photographer and imagine how you would take the picture.

• Content Type - What medium will the image be “printed” on. Ideas are photograph, painting, sketch etc

• Description - The main element of the image. Here be descriptive and details about the subject, it’s attributes & the scene

• Filter - Think image filters so describe the lighting or art style

• Composition - This covers both the perspective of the “camera” to the scene and final size of the image

• Enhance - Don’t forget to iterate, if it doesn’t look right the first time start changing the above attributes and try again

A photograph of a bear drinking from a river using a hyper realistic style. The image should be captured through the trees and zoomed in on the bear, 16:9 aspect ratio

Head over to Bright Journey AI for more technical articles.

💫Help Support my Work💫

I really enjoy researching and writing about AI & appreciate your support. If you enjoy this content please use any of the options below to help

• Buy Me a Coffee ☕️- AI is really fuelled by coffee - keep the tank filled

• Spread the Word 🗞️- Share on social or directly with your friends

• Give me Feedback 💬 - Tell me what’s important to you

• Follow me on X - Always happy to chat AI and software dev

📰 News 📰 

Google Unveils Gemma 2 for Advanced AI Applications

Google has launched Gemma 2, an advanced open large language model (LLM) available in 9 billion and 27 billion parameter versions. This model introduces significant improvements, including sliding window attention, soft-capping, and knowledge distillation, enhancing performance and efficiency. Gemma 2 supports diverse AI tasks, making it ideal for dialogue applications, natural language processing, and more. It integrates seamlessly with major AI frameworks like Hugging Face, NVIDIA, and Ollama, facilitating easy development and deployment. Additionally, Google is making Gemma 2 accessible through platforms like Google AI Studio and Hugging Face, enabling a broad range of AI innovations and research opportunities.

Developers can leverage Gemma 2 for various applications, including conversational agents, content generation, and data analysis. Its improved architecture promises better context understanding and response generation, setting a new standard for open AI models. By combining cutting-edge techniques and broad accessibility, Google aims to foster a more collaborative and advanced AI development environment.

Read More At: Hugging Face - Blog

OpenAI Introduces CriticGPT to Enhance AI Evaluation

OpenAI has developed CriticGPT, an AI model designed to critique and identify errors in responses generated by ChatGPT. This tool aids human trainers in spotting inaccuracies, enhancing the reinforcement learning process. CriticGPT has proven effective, helping trainers outperform those working alone 60% of the time. It was trained using a method where AI trainers inserted deliberate mistakes into code for CriticGPT to identify. This approach aims to improve AI alignment by ensuring more accurate and comprehensive feedback.

CriticGPT not only highlights potential errors but also suggests corrections, facilitating a more efficient and reliable model training process. OpenAI's initiative underscores the importance of continuous improvement and accuracy in AI development, aiming to create more dependable and trustworthy AI systems. This self-critique mechanism reflects a significant advancement in AI training methodologies, potentially setting new standards for the industry.

Read More At: OpenAI Blog

Toys "R" Us Unveils First AI-Generated Commercial Using OpenAI's Sora

Toys "R" Us has launched its first-ever AI-generated commercial, created using OpenAI’s Sora. The ad, titled "The Origin of Toys 'R' Us," was produced in collaboration with the creative agency Native Foreign and premiered at the 2024 Cannes Lions Festival. The project honours the legacy of the brand’s founder, Charles Lazarus, by showcasing cutting-edge AI technology to tell his visionary story​

Sora, a text-to-video tool, enabled the rapid production of the commercial, condensing hundreds of iterative shots into a polished final product. While the technology promises exciting creative possibilities, it has also sparked debate about potential job losses in the industry. Despite the controversy, the ad reflects a new era for Toys "R" Us, combining nostalgia with innovation to engage a modern audience​.

Read More At: Ars Technica

YouTube Negotiating with Labels for AI Music Generation

YouTube is in discussions with major record labels to license music for AI song generators, facing resistance from artists concerned about the devaluation of their work. The company aims to expand its AI tools and sign up more artists, offering upfront payments to labels. This move comes amidst legal battles with AI start-ups using copyrighted recordings. Record companies are embracing AI to create music and secure compensation, with Sony, Warner, and Universal in talks with YouTube. The industry is navigating the balance between innovation and protecting artists' rights.

Read More At: Ars Technica

🔨 AI Powered Tools 🔨 

• Midjourney - Platform that generates images from textual descriptions, catering to artists, designers, and creative professionals. It uses advanced machine learning to produce high-quality, diverse visuals based on user input, enabling seamless visualization of concepts and ideas.

• Visual Sitemaps - Tool designed for web developers and UX/UI designers. It automates the process of creating visual sitemaps of websites, helping to plan and audit site architecture efficiently. Utilizing AI, it captures and organizes screenshots of web pages to provide clear, navigable visual maps.

• Listen to Anything - ElevenLabs offers a text-to-speech solution that turns written content into high-quality, lifelike audio. It serves various industries by providing natural-sounding voice narrations for articles, PDFs, and other text formats, leveraging AI to enhance accessibility and content engagement.

🗞️ Other News Rollup 🗞️ 

AI in Software Development - Revolutionary AI tools like Devin AI and Codiumate are reshaping software development, emphasizing the need for human oversight.

AI in Biological Research - DeepMind's AI innovations democratize research by predicting protein structures and missense mutations, aiding disease studies.

AI Efficiency Breakthrough - Researchers eliminate matrix multiplication in AI models, reducing power consumption and improving sustainability for resource-constrained hardware.

AI in IBM Db2 - IBM upgrades Db2 with AI query optimizer, cluster management, and multi-tenancy support for improved performance and features.

AI Evaluation Evolution - Hugging Face upgrades Open LLM Leaderboard, LMSYS Chatbot Arena offers real-world evaluations, shaping AI landscape.

Metis AI Chatbot - Amazon developing advanced AI chatbot Metis for September release, facing competition from OpenAI and Google.

Datacenter Speed Boost - Intel introduces optical chiplet with 4 Tbps speed for AI and HPC, promising high efficiency and low latency.

Ethical AI for Government: Anthropic's Approach - Anthropic offers ethical AI models for government tasks like combating human trafficking, positioning itself as an ethical choice.

Innovative Datacenter Network - Alibaba Cloud unveils cutting-edge network design with GPUs, single-chip switches, and custom heat sinks.

Sohu ASIC Innovation - Etched's Sohu ASIC promises 20x performance, targets large batch sizes for efficient AI inferencing, secures $120M funding.

Stability AI Investment - Stability AI receives investments for text-to-image products, faces financial hurdles despite notable backers.

AI Country Restrictions - OpenAI to block users from unsupported countries like China, impacting developers and cloud platforms.

Enhancing AI Reliability - Google teams up with data providers to improve AI accuracy and introduces high-fidelity grounding for better performance.

Gemini AI Models - Google Cloud unveils Gemini 1.5 Flash and Pro, boosting AI capabilities with new features for developers.

Advanced Image Generation Model - Google's Imagen 3 on Vertex AI offers faster image generation and realistic rendering, benefiting companies like Shutterstock.

Vertex AI Upgrade - Google boosts Vertex AI with Mistral models, expanding developer options beyond Notebooks. Partnership reinforces Mistral's AI credibility.

🎶 Prompts 🎶 

Prompts used to generate some of this issues images. Unless otherwise stated all images are using Dall-E and ChatGPT Plus.

Thank You for Subscribing

Enjoying what you’re reading? Help me get better so I can continue to provide you with the most relevant content.