Sunday Rundown #76: Google Stuff & Diner Bear

Sunday Bonus #36: My quirky brainstormer bot

Happy Sunday, friends!

Welcome back to the weekly look at generative AI that covers the following:

• Sunday Rundown (free): this week’s AI news + a fun AI fail.

• Sunday Bonus (paid): a goodie for my paid subscribers.

Let’s get to it.

🗞️ AI news

Here are this week’s AI developments.

👩‍💻 AI releases

New stuff you can try right now:

1. Anthropic now has a “prompt improver” in its Console that can automatically improve your existing prompts using prompt engineering techniques.

2. ElevenLabs overhauled and added new features to its “Project” tool, which lets you turn long-form text into audiobooks or spoken narratives.

3. Google released quite a few things:

   1. The newest version of Gemini with the catchy name gemini-exp-1114 is sitting as #1 on Chatbot Arena, narrowly edging out the latest GPT-4o. (Try it on Google AI Studio.)

   2. Learn About is a new tool for structured research. It’s a bit like NotebookLM but with web access. It creates on-demand learning modules for your chosen topic, lets you chat about them, and can discuss uploaded images.
      
      

      Patrick Jordan
      (
      Tech & Nonsense
      ) and I had a 15-minute chat about it yesterday, in case my talking face is something you care to listen to:

   3. The previously announced Vids tool that turns text prompts into video slide decks is now generally available:

4. OpenAI made the Windows ChatGPT app available to everyone while giving the Mac version the ability to read and interact with content from your coding apps.

5. Vidu released V1.5 of its video model, which can incorporate multiple image references into the generated videos, along with other big upgrades. Vidu V1 was already in the “positive surprises” tier in my image-to-video test. (Try it here.)

---

🔬 AI research

Cool stuff you might get to try one day:

1. Google and the National University of Singapore previewed a tool called ReCapture that can change the camera movements for an existing video.

2. Researchers in China teased MagicQuill, an image-editing tool that automatically predicts your intended edits without you having to describe them using text:

3. OpenAI is reportedly planning a January release of its AI agent (dubbed “Operator”) that can take actions on your computer (a la Anthropic’s “Computer Use.”)

4. Suno teased the imminent release of V4 of its AI text-to-music model. (I pitted Suno V3 against Udio back in April.)

5. YouTube will let creators restyle select licensed songs for their YouTube Shorts using AI. (Currently in testing with a small group of creators.)

---

📖 AI resources

Helpful AI tools and stuff that teaches you about AI:

1. “How 200 real-world businesses are transforming with AI” - a roundup of case studies by Microsoft.

2. “How Notion Cofounder Simon Last Builds AI for Millions of Users” [VIDEO] - another great practical chat by Dan Shipper of Every.

3. “New Google Model Ranked ‘No. 1 LLM’, But There’s a Problem” - another great dive by AI Explained:

4. “Student’s Guide to Writing with ChatGPT” - handy tips for students by OpenAI.

5. “State of Data + AI” [PDF] - a deep dive by Databricks.

6. “The Fall 2024 Workforce Index” - a global survey of 17,000 desk workers by Slack.

7. Voice Lab - an open-source framework to evaluate LLM-powered agents by Substack’s own

   Sahar Mor
   . (If you aren’t already, follow his
   AI Tidbits
   newsletter.)

8. “Where’s the Value in AI?” [PDF] - a free report by the Boston Consulting Group

---

🔀 AI random

Other notable AI stories of the week:

1. After three months away, Greg Brockman is back at OpenAI.

2. Perplexity is starting to experiment with ads and sponsored content in the US. (Something I pointed out was coming in last week’s “AI Search” article.)

🤦‍♂️ AI fail of the week

Man, 1950s ads were crazy! (One of DALL-E’s extra attempts from this test.)

---

💰 Sunday Bonus #36: Get creative with my quirky brainstormer bot

We all know that LLMs can be great at helping you brainstorm ideas.

But I found that giving them constraints often results in better, more creative responses. It’s a bit like forcing chatbots into lateral thinking.

So I put together and fine-tuned a fun Poe bot that helps you think outside the box when coming up with ideas for a topic of your choice.

It uses the “Ask me questions” method to understand your request better and suggests topic-appropriate constraints to make exploration more fun and unconventional.

Check it out!