ChatGPT Can Copy Your Voice

PLUS: CodexGraph for LLMs, Spreadsheet AI Agents

The fastest way to build AI apps

Writer is the full-stack generative AI platform for enterprises. Quickly and easily build and deploy AI apps with Writer AI Studio, a suite of developer tools fully integrated with our LLMs, graph-based RAG, AI guardrails, and more.

Use Writer Framework to build Python AI apps with drag-and-drop UI creation, our API and SDKs to integrate AI into your existing codebase, or intuitive no-code tools for business users.

Today’s top AI Highlights:

  1. CodexGraph for LLMs to better understand large codebases

  2. Workflows by LlamaIndex for a new way to build complex AI apps

  3. ChatGPT’s Advanced Voice Mode can imitate your voice

  4. AI model that can listen while speaking

  5. AI agents to fill your spreadsheets and automate research

& so much more!

Read time: 3 mins

Latest Developments

LLMs have proven their prowess in tackling individual coding tasks, but they often stumble when faced with the complexities of entire code repositories. Introducing CodexGraph: it transforms code repositories into graph databases. This allows LLMs to query and navigate the codebase more effectively. By representing code elements and their relationships as nodes and edges, CodexGraph provides LLMs with a structured understanding of the code.

Key Highlights:

  1. Graph-Based Representation - CodexGraph translates code into a graph database, representing code elements as nodes and their relationships as edges. This provides a structured and intuitive representation for LLMs to understand.

  2. Code Retrieval with Graph Queries - Instead of relying on imprecise similarity searches, CodexGraph allows LLMs to use a graph query language to precisely locate specific code snippets and structures within the repository for more accurate code retrieval.

  3. Validated Performance on Benchmarks - Testing on three code-related benchmarks: CrossCodeEval, SWE-bench, and EvoCodeBench, it demonstrated its ability to handle complex code-related tasks.

  4. Practical Applications - CodexGraph has been successfully applied as AI agents for 5 real-world coding applications, including code debugging, unit test generation, and code documentation. Here’s a demonstration you can check out.

Tired of rigid, linear AI application development? LlamaIndex's new “Workflows” feature in beta lets you build dynamic, responsive AI applications that can handle complex logic with ease. Think of it like a conductor for your AI orchestra: it directs different components (like LLMs, data loaders, and vector databases) to work together seamlessly. Forget clunky DAGs and query pipelines – Workflows offer a more intuitive, event-driven approach that allows for loops, error handling, and state management, giving you the power to build truly intelligent and adaptable AI applications.

Key Highlights:

  1. Adaptive Logic with Loops and State - Workflows empower your AI applications to learn and adapt. They can execute loops to refine results, handle errors gracefully, and even maintain a memory of past interactions through a shared context, making your applications more intelligent and responsive.

  2. Mix-and-Match Components with Ease - Effortlessly plug in different AI building blocks like Legos. Workflows provide a flexible framework where you can combine pre-built components or create your own custom steps to control your application's behavior.

  3. Debug Like a Pro - Workflows comes equipped with powerful visualization and debugging tools. See the flow of events, pinpoint bottlenecks, and understand exactly how your application is working with just a few clicks. 

Quick Bites

  1. OpenAI released their GPT-4o model’s system card, including the Advanced Voice Mode, on Thursday, detailing the safety measures they have undertaken before releasing the models. The report says the Advanced Voice Mode can imitate a user’s voice in rare instances, and they are implementing safeguards to prevent this.

    Listen to this audio clip where the model outbursts “No!” then begins continuing the sentence in a similar sounding voice to the red teamer’s voice

  1. AI researchers at ByteDance have developed an AI model that can listen while speaking, a crucial aspect of human conversations. This “listening-while-speaking” model improves conversational AI by allowing seamless, duplex communication and handling interruptions during conversations.

  2. The little orange AI device Rabbit R1 has received some updates. A new “beta rabbit” mode adds conversational AI chops to the device. Other basic features like alarms and timers have been improved. But the much-anticipated Large Action Model is still under wraps.

  3. OpenAI has appointed Zico Kolter, a Carnegie Mellon professor specializing in AI safety and alignment, to its board of directors. Kolter has also joined the Safety and Security Committee to bring his expertise to the board.

  4. ChatGPT Free users can now generate up to two images per day with DALL·E 3 within ChatGPT.

Tools of the Trade

  1. Matrices: Automates research by filling spreadsheets with data from multiple sources, including proprietary information. It simplifies complex tasks using natural language and custom automation.

  1. Recap by Fabric: Get personalized AI summaries, in your inbox, of everything you’ve saved, created, or captured, helping you revisit and reflect on your content.

  2. Tara AI: Helps software teams optimize performance by connecting issue tracking and Git source control for real-time insights and alerts. It simplifies project management for better progress tracking and resolving blockers.

  3. 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. Advanced Voice on GPT-4o needs custom instructions/GPTs. Not only would it help with many use cases (education, feedback on performances, even games) but I really want it to talk faster by default and differentiate between chit-chat and informational requests. Soon I suspect. ~
    Ethan Mollick

  2. however strong you think gpt-5 pushback will be, it will be stronger. however well-organized you think anti-ai opposition will be, it'll be more cohesive.
    it's gonna be clear to 70% of humanity soon that they no longer have a job. ~
    Aidan McLau

Meme of the Day

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!

Unwind AI - Twitter | LinkedIn | Instagram | Facebook

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 at least one (or 20) of your friends!

Reply

or to participate.