Welcome, AI RESEARCHERS.

Instagram is testing a feature that could change how influencers interact with followers. Soon, your favorite creator might have an AI twin answering your messages.

Instagram's 'AI Studio' allows creators to generate AI versions of themselves, primarily for messaging followers and answering common questions. Learn more..

In today’s AI Researchs:

• Instagram Launches AI Chatbots for Creators

• Time Magazine Partners with OpenAI and ElevenLabs

• OpenAI Boosting AI Accuracy with CriticGPT

• Gemma 2: New AI advanced Model Released

• Etched AI Unveils Sohu

• Automating Blogging with AI: Step-by-Step Guide

• AI Researchs & AI pointers

• Trending Ai tools

• More Ai quick hits

Read time: 6 minutes

AI creator

📱Instagram Launches AI Chatbots for Creators

AI Researchs: Instagram is testing "AI Studio," allowing creators to make AI versions of themselves for fan interactions. CEO Mark Zuckerberg announced this early test in the US, with AI chatbots appearing primarily in messaging.

For more details:

• Meta's CEO Mark Zuckerberg announced that AI Studio is currently in an early test phase in the US. Users might start seeing AI versions of their favorite creators on Instagram, primarily in messaging, with clear labels indicating AI generation.

• Messages from these AI chatbots include warnings about potential inaccuracies or inappropriate content, marked with AI and beta tags for transparency.

• Meta collaborated closely with creators to develop these chatbots, aiming to make interactions both helpful and entertaining.

• Zuckerberg also mentioned the potential for small businesses and individuals to create their own AIs, envisioning dynamic and personalized interactions.

• This initiative aligns with Meta’s long-term vision of integrating AI assistants and chatbots into user interactions, a concept previously discussed in company announcements.

Why this is important: Instagram’s AI Studio lets creators make AI versions of themselves to answer followers' questions and engage more dynamically. If successful, this could change ecommerce support, customer support and how creators interact with their audience, offering more personalized and immediate responses.

AI PARTNERSHIPS

🤝 Time Magazine Partners with OpenAI and ElevenLabs

AI Researchs: Time Magazine, a century-old publication, is stepping into the AI era through new collaborations with OpenAI and ElevenLabs, aiming to enhance content distribution and introduce AI-powered audio features.

For more details:

• Time is giving OpenAI access to its current and past content for AI training. In exchange, ChatGPT will provide summaries of Time articles with links back to the original content.

• This is OpenAI's eighth known partnership with a media company, following deals with The Atlantic, Vox Media, and others.

• Time will use ElevenLabs' "Audio Native" player on its website. This tool uses AI voices to automatically read articles aloud. It has been tested since 2023 and is now active on some articles.

• Time will get access to OpenAI's new tools and models for its journalism work. The magazine will also give feedback on how AI affects news presentation.

• The ElevenLabs partnership aims to make Time's content available in audio form, potentially reaching more people.

Why this is important: Time Magazine's partnerships with OpenAI and ElevenLabs bring AI tools to journalism, potentially broadening their audience through new content formats and innovative delivery methods. Whether these AI efforts will succeed remains to be seen, but they offer immediate utility and the possibility of a more engaged readership.

OPENAI ANNOUNCEMENT

⚗️ OpenAI Boosting AI Accuracy with CriticGPT

AI Researchs: CriticGPT, a model derived from GPT-4, critiques ChatGPT responses to help human trainers spot mistakes, improving the overall training process.

For more details:

• CriticGPT is a specialized model based on GPT-4 designed to identify errors in ChatGPT’s outputs. It assists human trainers by highlighting inaccuracies, making it easier to refine and improve AI performance.

• Human trainers using CriticGPT identify mistakes 60% more often than those working without it.

  Training Approach: CriticGPT was trained on ChatGPT's incorrect outputs. Trainers inserted deliberate mistakes for CriticGPT to critique, helping it learn to spot errors effectively.

• Combining human efforts with CriticGPT leads to more thorough and accurate critiques, with fewer false positives, compared to reviews done without AI assistance.

• In over 60% of cases, trainers preferred critiques from the Human+CriticGPT team over those from unassisted trainers, citing clearer and more precise feedback.

• The next steps include addressing complex and dispersed errors and developing tools to help trainers handle sophisticated AI outputs better.

Why this is important: As AI systems grow more complex, tools like CriticGPT are crucial for helping trainers accurately evaluate outputs, ensuring the reliability and safety of AI technologies.

GEMMA 2

🚨 Gemma 2: New AI advanced Model Released

AI Researchs: Google DeepMind has introduced Gemma 2, a powerful and efficient AI model available to researchers and developers in 9B and 27B parameter sizes.

For more details:

• The 27B model performs better than models over twice its size, and the 9B model outperforms similar models like Llama 3 8B. Both offer excellent results for various AI tasks.

• Gemma 2 can run on a single NVIDIA H100 Tensor Core GPU or TPU host, significantly lowering deployment costs while maintaining high performance.

• Optimized for fast inference across various hardware setups, including laptops, desktops, and cloud systems, ensuring versatile and quick AI applications.

• Supports a context length of up to 2 million tokens, enabling the handling of more extensive and complex tasks.

Why this is important: Gemma 2's open, high-performing AI models are now available to more developers and researchers, making advanced AI tools more accessible and affordable, which can accelerate AI innovation and development.

big Ai startup

🌟 Etched AI Unveils Sohu : AI makes the biggest bet in AI chips

AI Researchs: Etched AI, founded by Harvard dropouts, is developing the Sohu chip - an AI chip designed specifically for transformer models, challenging the dominance of general-purpose GPUs in the AI hardware market.

For more details:

• Etched AI's Sohu chip is an application-specific integrated circuit (ASIC) focused solely on running transformer models, which are widely used in generative AI applications like text, image, and video generation.

• Built using TSMC's 4nm process, the Sohu chip claims to offer superior performance and energy efficiency compared to traditional GPUs. The company states that one Sohu server can replace 160 Nvidia H100 GPUs.

• The startup is betting on the continued scaling of transformer models in AI, arguing that as models grow larger and more complex, specialized chips will become necessary to handle them efficiently

• Etched AI recently closed a $120 million Series A funding round, bringing their total raised to $125.36 million. High-profile investors include Peter Thiel and GitHub CEO Thomas Dohmke.

Why this is important: The company aims to make AI models more accessible and efficient, potentially enabling real-time video, audio, agents, and search capabilities that are currently limited by hardware constraints.

AI tutorial

✅Automating Blogging with AI: Step-by-Step Guide 

AI Researchs: Learn how to set up an AI-automated blogging system to generate and post blog content, potentially earning up to $30,000 per year.

Step by step

Step 1: Gather News Articles

1. Google News: Go to news.google.com.

2. Search and Select: Find an article relevant to your blog and copy its URL.

Step 2: Create an RSS Feed

1. Sign Up on RSS.app: Register at RSS.app.

2. Generate Feed:

   • Create a new feed using the copied article URL.

   • Save the generated RSS feed URL.

Step 3: Set Up Automation with Make.com

1. Sign Up: Create an account at make.com.

2. Create Scenario:

   • Add "RSS" as the trigger module.

   • Enter the RSS feed URL.

   • Set the frequency for checking new articles.

Step 4: Generate Blog Content

1. Add OpenAI GPT-4 Module:

   • Use your OpenAI API key to authenticate.

2. Configure Content Generation:

   • Use a prompt to create blog posts from the RSS feed data.

   • Example prompt: “Generate a blog post based on this article: [article content]. Make it SEO-friendly and match the tone of [your blog's name].”

Step 5: Automate Posting to WordPress

1. Connect to WordPress:

   • Add a WordPress module in Make.com.

   • Provide your WordPress site details and API key.

2. Configure Post Settings:

   • Set the title, content (from GPT-4 output), and other details.

   • Choose immediate publishing or schedule the post.

AI Research Roundup

The study of 300 tech MSMEs in five Indian cities shows widespread interest in AI. While 94% see AI's potential for growth, 72% say they need AI training, and 59% mention money issues as barriers to using AI.

Researchers have developed a new AI framework, agent symbolic learning, treats AI agents as symbolic networks optimized by natural language prompts and tools. This method allows AI agents to autonomously evolve and improve post-deployment, showing promising results in both standard benchmarks and real-world applications.

AI Pointers

Ray kurzweil says in the 2030s, nanobots will connect our brains to the cloud, expanding intelligence a million fold by 2045, allowing us to be funnier, sexier, smarter, more creative, and free from biological limitations. We'll choose our parents, visualize objects in 11 dimensions, speak all languages, expand consciousness, and experience richer culture with AI enhancing our lives, giving it the greatest meaning.

Larry Summers says studying coding is a bad idea and instead young people should learn to work creatively in a group and define purpose, because AI is coming for IQ before it comes for EQ

Mark Zuckerberg says some in the tech industry are trying to build one true AI God, but the future will be more like a lot of AIs

• ✨ lexica - AI generated art gallery - See prompts others have used.

• 📈 June - AI-driven customer insights tailored for product-oriented teams.

• 📲 Pygma - Your AI assistant for managing social media.

• 🎨 Figma AI - Integrated AI in Figma that converts text into visual designs effortlessly.

• 📞 Character AI Call - Engage in phone conversations with AI characters.

• 💬 Sider ChatScreen - AI chatting for on-screen content, accessible on iOS.

• ❓Question Base - Auto-respond to repeated questions in Slack.

• 🏋‍♂ The Fittest - Use AI to make your fitness routine more engaging.

• 🖍 AI Gradient Generator - Instantly generate gradients from text prompts

• 💡 TextMine - Quickly analyze, manage, and search thousands of documents for informed decisions.

• 🔄 Assign-AI - Automate workflows easily, no coding required.

• 🎨 Image Generator by Leap - Produce high-quality visuals from text prompts to improve your projects.

• 💻 FreeAIChatbot - Access Claude 3.5, GPT-4, Llama 70B, and more from a single platform.

OpenAI postpones ChatGPT Voice Mode alpha release by one month, developing content detection, user interface, and scalability; plans iterative deployment starting with a small user group, targeting fall availability for all Plus subscribers.

macOS users can now download the ChatGPT desktop app, offering quicker access to ChatGPT via the Option + Space shortcut for discussing emails, screenshots, and screen content.

Video editing app Captions has launched an AI edit feature that automatically adds graphics, effects, transitions and more to vertical talking head videos

SoftBank's Vision Fund 2 will invest $10-$20 million in Perplexity AI, valuing this search startup at $3 billion as part of a $250 million funding round.

Character.ai introduces Character Calls, enabling users to engage in voice conversations with AI characters via their mobile app for tasks ranging from language learning to creative storytelling.

ElevenLabs introduces the Reader App, a mobile application that converts text to speech using high-quality voices, enabling users to listen to any text content while on the move.

Google Translate has expanded its language support by adding 110 new languages, bringing the total to 243, as part of its initiative to cover 1,000 languages globally using AI technology

Major Music industries sues AI music generators Suno and Udio for alleged copyright infringement, seeking $150,000 per song used in training data.

YouTube's plan to license music from major labels for AI tools comes amid industry lawsuits against unlicensed AI music generators

Stability AI secures $80 million funding round and appoints new CEO Prem Akkaraju to revitalize company.