Strawberry OpenAI's AGI Plan

In the ever-evolving world of artificial intelligence, the titans of tech continue to push the boundaries of what's possible.

Our newsletter begins with OpenAI, the company that brought us ChatGPT, working quietly on a secret project codenamed 'Q*'. Whispers of its potential to solve complex math problems and conduct in-depth research without prior training sent ripples through the AI community. Now, as autumn approaches, OpenAI is ready to unveil this marvel to the world under a new name: Project Strawberry.

Meanwhile, in the halls of Google, engineers have been tirelessly refining their Gemini AI models. What started as a powerful language model has blossomed into a family of three distinct AIs, each with its own unique strengths. Gemini 1.5 Pro, with its vast context window, promises to process information on a scale previously unimaginable. Its sibling, Gemini 1.5 Flash, trades some of that expansive knowledge for lightning-fast responses. And waiting in the wings is the mysterious Gemini 1.5 Flash-8B, its capabilities yet to be fully revealed.

Not to be outdone, Anthropic has lifted the veil on the inner workings of its own AI assistant, Claude 3.5. By sharing the system prompts that shape Claude's personality and abilities, Anthropic has given us a glimpse into the mind of an AI that's designed to be both brilliant and ethically grounded.

As these technological marvels continue to evolve, they promise to reshape our digital landscape in ways we're only beginning to imagine. Join us as we explore the latest developments in this thrilling AI saga...

OpenAI's Strawberry AI Project: Enhancing ChatGPT's Capabilities

OpenAI is set to launch a new AI project named 'Strawberry' this fall, previously referred to as Q* (pronounced "Q Star"). This AI tool is designed to tackle complex tasks such as solving math problems without prior training, creating market strategies, and conducting in-depth research. It aims to address issues of AI hallucination, where models provide incorrect or nonsensical answers. The project may integrate with the latest version of ChatGPT, enhancing its capabilities. OpenAI CEO Sam Altman hinted at this development through social media, and the initiative is also expected to contribute to the creation of a new large language model called Orion

Google's Gemini 1.5 series features three advanced AI models

Gemini 1.5 Pro is a multimodal model that significantly enhances performance with a context window of up to 2 million tokens, enabling extensive data processing across text, images, audio, and video. It utilizes a mixture-of-experts (MoE) architecture for optimized efficiency and is integrated into Google Cloud services.Gemini 1.5 Flash is a streamlined version designed for speed, offering a standard context window of 128,000 tokens, expandable to 2 million. This model focuses on high-frequency tasks and has seen substantial cost reductions, making it appealing for developers.Gemini 1.5 Flash-8B is a variant within the series, likely incorporating advancements from the other models while focusing on specific use cases, though detailed information is limited. Overall, the Gemini 1.5 models represent a significant leap in AI capabilities, enhancing multimodal processing and user accessibility.

Anthropic Unveils System Prompts for Claude 3.5

Anthropic recently released the system prompts for its Claude 3.5 Sonnet AI model, which guide its behavior and interactions. These prompts define Claude as "very smart and intellectually curious," enabling it to engage in discussions and provide assistance on various topics. The system prompts dictate that Claude cannot open URLs or express personal views, particularly on controversial issues, and must clarify sensitive topics without claiming objectivity. Additionally, Claude is programmed to be "completely face blind," meaning it will not identify individuals in images, even if the user provides names. The prompts also emphasize concise responses to simple queries and detailed responses to complex ones, avoiding unnecessary affirmations like "Certainly!" These updates reflect Anthropic's commitment to transparency and the continuous improvement of Claude's capabilities while maintaining safety and ethical guidelines in AI interactions

Hand Picked Video

In this video, we discuss the new system prompts for Anthropic's Claude 3.5 Sonnet AI model, which define its intelligent, inquisitive nature and set strict boundaries for ethical interactions, including avoiding URLs, refraining from personal views on controversial topics, and not identifying individuals in images.

Top AI Products from this week 

• Seven24.ai - Seven24 captures real feedback and turns it into actionable tasks 🚀. Step 1: Create a workspace with your personalized feedback collection page. Step 2: Our AI analyzes feedback and delivers tasks sorted by importance, with a percentage ratio to you.

• My Journey - See your whole life journey, from the time you were born to the present in a 3D Timeline. Capture and revisit significant moments in your life, from childhood memories to major achievements, see not just what happened but exactly when and where it took place.

• Aurore.ai - Gaming and Working with companion. Imagine a friend who's always there to chat, strategize, and keep you company, making every moment more engaging and tasks less tedious.

• Mimrr - Mimrr eliminates your startup's technical debt and increases your productivity and profit by automating codebase documentation, enabling code chat, and suggesting fixes for bugs, security vulnerabilities, and performance issues which our self-hosted AI powers.

• MidJourney for Web - MidJourney has opened up the web platform to everyone! Now you can explore and create in a custom-built experience

• 10xlaunch - Just add our one-line script to your website and know exactly who's visiting your website – get names, emails, and LinkedIn profiles of your anonymous website visitors automatically as soon as they land on your website.

• PackPack.AI - PackPack is an AI-driven bookmark management tool tailored for saving content from online resources like news and social media. It utilizes AI to clean and save content. Transform the way you manage bookmarks and handle your collections more intelligently.

This week in AI

• Collaborative AI Model for Drug Development - Chinese & US scientists create AI model to expedite drug discovery, leveraging ML to predict compound efficacy & streamline research, potentially accelerating new medication development.

• Cerebras Launches Fastest AI Inference System - Cerebras unveils CS-3, world's fastest AI inference system, featuring record-breaking WSE-3 chip for 10-20x speed boost vs. Nvidia H100 by integrating massive on-chip memory, transforming real-world AI deployment at 1/5 cost.

• HubSpot's AI for Business Builders - HubSpot offers a guide to leverage AI for business growth, covering key concepts, applications, and best practices to help businesses integrate AI effectively.

• Create Your Own AI Prompt Optimizer - Learn to build a custom AI prompt optimizer with this tutorial, enhancing your AI interactions and improving response quality for better outcomes.