BB Digest: Ship-mas begins

PLUS: Smart enough but cheap models

December 05, 2024

Digest #516 → Subscribe | Upgrade to Pro

Hey folks,

OpenAI says it’ll ship something everyday for the next 12 days (starting today). You can follow this live blog for each day’s updates if you don’t want to wait for these digests. I’m going back to scrolling Twitter like crazy :)

TLDR; inside today’s newsletter

  • My wishlist for OpenAI, Amazon’s third door to AI and Google’s side quests

  • ICYMI: 5 ways to save time in creating content

  • snack n skill - make visuals from your data (without Excel gymnastics)

  • 6 of the best new AI tools

  • 5 interesting posts from around the web

  • to-do’s

Alright, let’s get to it!

  • OpenAI’s been shipping just blog posts for a while now, and there are more GPT-4 like models than ever. Let’s see what they have in store but here are my expectations:

    • Full o1 model with access to files, search etc.

    • Sora, maybe a faster version called Turbo (with more coming later)

    • ChatGPT getting ‘eyes’ finally

    • More desktop upgrades (OAI’s version of computer use, MCP)

    • Image generation with GPT-4o

    • better memory, canvas, and more tools

    • GPT-4.5/GPT-Next? Agents? GPT store revival?

  • Amazon is finally in the AI game. It has launched a new set of foundational models - Amazon Nova. These are not the best-performing models but they are much cheaper than comparable champs out there. It is enough to keep AWS customers on their platform and other features might attract new business too.

  • Looking for funding for your startup? Forum Ventures invests in early-stage B2B SaaS companies and is making their final investments in 2024. Get a $100k plus highly personalized GTM, sales and fundraising support through their world-class accelerator. Don’t wait – pitch them today!*

  • Google also made a bunch of noise yesterday. Veo (Google’s answer to Sora) is now available to more customers in private preview. Imagen 3 is also generally available to Vertex AI customers now. Genie-2, a project from Google Deepmind can create playable games from single images (more on this below). Plus, GenCast, a new weather prediction model uses image-gen techniques to beat current prediction models.

    *sponsored

👀 ICYMI: New stuff we’ve launched

🍽️ Snack ‘n’ skill

Learn how to create interactive data visualizations with Claude

Manually creating charts and graphs from spreadsheets is time-consuming. It often requires coding knowledge or Excel expertise. With Claude's new Analysis Tool, you (even as a beginner) can instantly turn your data into insightful visualizations through simple conversation.

Step 1. Prepare your data

Start by getting your data into CSV format. If you don't have a CSV/Excel sheet ready, you can use ChatGPT to convert your existing documents (PDFs, docs, presentations) into CSV format.

Step 2. Create your visualization

  • Upload your CSV file to Claude using the paperclip icon

  • Ask Claude to visualize specific aspects of your data (e.g., "show me best-selling products")

  • Refine the visualization by asking Claude to modify the chart type, colors, or focus

  • Get both visual and text-based analysis of your data

  • Experiment with different views using Claude's suggested prompts

  • Publish your visualization as a shareable artifact that others can view

Tip: When asking Claude to modify visualizations, be specific about what insights you're looking for. Instead of "show me sales data," try "visualize monthly sales trends by product category."

This walkthrough is a condensed version of this tutorial. Check it out for step-by-step instructions and example prompts for creating beautiful data visualizations.

⚙️ Top new tools

  • Exa Websets - Find a comprehensive set of anything you want from the web. (waitlist, but worth signing up)

  • Conversation AI by Elevenlabs - Add voice to your agents on web, mobile or telephony in minutes.

  • AI Santa - Free personalized videos from Santa.

  • Roster - Hiring platform for creators with AI-powered matchmaking.

  • Cloudy - Functional technical documentation that stays up-to-date with your code.

  • Plot - Unlock AI-powered consumer insights from social media videos.

🗣️ Everyone’s talking about…generating 3D worlds

Two major AI labs have unveiled systems to generate explorable 3D environments - Google DeepMind's Genie 2 and World Labs' world generation system. Both convert 2D images into interactive 3D spaces, but take different approaches to the challenge.

What's going on here?

AI can now turn single images into consistent, physics-based 3D environments that humans and AI agents can explore and interact with in real-time.

What does this mean?

World Labs has developed "Midjourney for 3D spaces," allowing users to upload an image and generate a 3D environment. Creators can stage scenes and plan camera movements easily, without complex 3D software, while the system ensures physics rules and spatial consistency.

DeepMind's Genie 2 acts as an AI game engine, generating interactive 3D worlds from image prompts, complete with physics, character movements, and NPCs, maintaining consistency for up to a minute for testing and prototyping.

Why should you care?

If you're in product design, marketing, or development, this is a game-changer for rapid prototyping. Instead of spending weeks in 3D software or paying thousands for visualization, you could test ideas in minutes. Want to see how your product looks in different environments? Need to visualize a store layout? Plan a video shoot? All possible in real-time now.

 📜 Interesting posts

📌 To-dos

  1. Let us know if you have any tutorials or courses you want us to create—we’re always open to ideas.

That wraps things up for today! See you again next week and on the live blog 👋

Ben

Enjoy this newsletter? Please forward to a friend.

Want more education on AI? Become a Ben’s Bites member and access our courses & tutorials.

Want to advertise in this newsletter? Click here.

Reply

or to participate.