Today’s top AI Highlights:

1. Google releases opensource Gemma 2 2B and other AI tools

2. Understand your customers’ voice, just ask AI

3. Cohere releases AI Prompt Tuner to generate optimal prompts

4. Runway releases 7x faster Gen-3 Alpha Turbo video AI model

5. Opensource alternative to OpenAI’s SearchGPT

& so much more!

Read time: 3 mins

Latest Developments

Google Expands Gemma Family with a Small Model 👨‍👨‍👦

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.

Key Highlights:

1. Gemma 2 2B - This lightweight model delivers exceptional performance for its size, and optimized for deployment on a wide range of hardware. It beats GPT-3.5 and Mixtral 8× 7B, models 10x of its size in the Chat Arena.

2. ShieldGemma - ShieldGemma offers state-of-the-art classifiers that detect hate speech, harassment, sexually explicit content, and dangerous content.

3. Gemma Scope - Gemma Scope is an open suite of Sparse Autoencoders for Gemma 2 9B and 2B. Sparse Autoencoders are like "microscopes" that can help us break down a model’s internal activations into the underlying concepts.

4. Open and Accessible - You can get the Gemma 2 2B model on the Hugging Face. You can also try Gemma 2 2B in Google AI Studio.

Stop Searching, Start Understanding What Customers are Saying 😎

We've all been there – drowning in customer feedback, struggling to pull out meaningful insights quickly. Especially when different teams require different data to make decisions, it can feel like you're constantly digging through data. Syncly AI just made it much easier to get everyone on the same page with their new AI Search. Now, anyone can get the customer insight they need in seconds.

Key Highlights:

1. Just Ask - Instead of running queries all day long, simply type in a question like "What are customers saying about our new refund policy this month?”

2. Get instant, AI-powered summaries - Syncly analyzes the results and provides a concise overview of key trends, sentiment, and top 10 issues, saving you valuable time.

3. Easy sharing for better collaboration - Instantly share search results with a single link, ensuring everyone from product and customer teams to leadership has access to the vivid voice of the customer.

Ready to see Syncly in action? Sign up for a demo and get a free analysis of your critical customer issues.

Cohere’s New AI Prompt Tuner to Opmize Prompts 👩‍🔧

Getting the perfect response from an AI model depends entirely on the prompt. AI engineers spend 20-30% of their time on prompt engineering. To make the process more efficient and robust, Cohere has introduced Prompt Tuner in beta. Prompt Tuner is your personal prompt optimizer that uses a powerful evaluation and refinement loop to improve your results. The best part is you can define what “success” looks like!

Key Highlights:

1. Customizable success criteria - You can define what constitutes a successful prompt based on your specific needs. This could include metrics like accuracy, character count limits, adherence to JSON format, grounding, and more.

2. Iterative refinement process - It employs an iterative approach to optimization. It generates completions using the provided prompt, evaluates them against the defined success criteria, and then uses this feedback to generate an improved prompt.

3. Integrated with Cohere Command R+ - Prompt Tuner uses Command R+ model which is strong in instruction-following, to generate the improved prompts during each iteration. This ensures that the refined prompts guide the LLM toward the desired outputs.

Quick Bites

1. 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.

2. Following the OpenAI-Scarlett Johansson debacle over the new Voice Mode, senators are introducing the No Fakes Act to give individuals the federal right to approve the use of their voice, appearance, or likeness in AI replicas. This bipartisan legislation seeks to prevent unauthorized AI-generated imitations.

3. AMD is becoming an AI chip company like NVIDIA. They just released their Q2 2024 earnings which shows a staggering 115% YoY rise in revenue from data centres. This growth is driven by their new MI300 AI accelerators that compete directly with NVIDIA H100. This AI chip just did over $1 billion in sales in a single quarter.

4. Meta got some mud on itself after its chatbot Meta AI entirely denied the existence of an assassination attempt on Donald Trump, while providing detailed information about Kamala Harris’s campaign.

   

   The Meme Lab, Billboard Chris, Jack Poso

   
   Meta has responded by saying that their AI was programmed to avoid potentially inaccurate information about breaking news, which caused the initial issues, and they are now working to correct these errors and improve their systems.

Tools of the Trade

1. OpenSearch GPT: Opensource AI search engine that personalizes results based on your interactions and preferences. It learns about you over time so you can control what it knows.

2. Not Diamond: AI model router that selects the best model for each task, and reduces costs and latency while maintaining high performance. It also supports custom router training with your data.

3. Llama Tutor: Your personal AI tutor built with Llama 3.1 running on Together AI. Enter a topic you want to learn about along with the education level to generate a personalized tutor.

4. Awesome LLM Apps: Build awesome LLM apps using RAG for interacting with data sources like GitHub, Gmail, PDFs, and YouTube videos through simple texts. These apps will let you retrieve information, engage in chat, and extract insights directly from content on these platforms.

Hot Takes

1. AI will be 100X more capable than GPT-4 by 2027, while compute costs are falling by 75-85% per year. This will TRANSFORM the world. ~
   Peter H. Diamandis
   

2. Intelligence is becoming so cheap that everyone can afford to be dumb now. ~
   Bojan Tunguz

That’s all for today! See you tomorrow with more such AI-filled content.

Real-time AI Updates 🚨

⚡️ Follow me on Twitter @Saboo_Shubham for lightning-fast AI updates and never miss what’s trending!

PS: We curate this AI newsletter every day for FREE, your support is what keeps me going. If you find value in what you read, share it with your friends by clicking the share button below!