Sunday Rundown #69: Meta Connect & Sexy Gollum

Sunday Bonus #29: Creating NotebookLM "Audio Overviews" without a source.

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. Google released two upgraded Gemini models—Gemini-1.5-Pro-002 and Gemini-1.5-Flash-002—that are simultaneously cheaper, faster, and better versions of their predecessors.

2. Google’s already awesome NotebookLM can now natively parse audio files and YouTube videos as sources.

3. Meta announced a slew of AI developments at Meta Connect:

   1. A new live voice model, which will eventually feature the voices of famous people.

   2. The ability to talk about and edit your photos in Meta AI using natural language.

   3. Automatic translations of the audio in your Reels.

   4. More AI tools for businesses.

   5. More AI-powered image and theme creation features in Meta products.

   6. New Llama 3.2 family of multimodal models.

4. OpenAI finally rolled out the Advanced Voice Mode in ChatGPT to Plus and Teams users…but not in the UK, EU, and a few other European countries. Classic.(There are also five new voices to pick from.)

5. Runway now lets you generate vertical videos using Gen-3 Alpha Turbo.

6. Suno Pro and Premier users can now crop songs to adjust their start and end times.

---

🔬 AI research

Cool stuff you might get to try one day:

1. Alibaba Group presented an impressive framework MIMO that can synthesize any reference image into an existing video.

2. Allen Institute for AI introduced a Molmo family of open multimodal models that are competitive with or outperform the best existing vision models.

3. Duolingo teased lots of AI-powered experiences at its Duocon 2024 event.

4. Microsoft launched a preview of a “Correction” feature in Azure AI that can allegedly detect and fix LLM hallucinations in real time.

---

📖 AI resources

Helpful stuff that teaches you about AI:

1. 185 real-world gen AI use cases from the world's leading organizations - a collection by Google.

2. AI Data Drop: 3 Key Insights from Real-World Research on AI Usage - takeaways from Copilot usage at 58 companies by Microsoft.

3. ‘Advanced Voice’ ChatGPT Just Happened … But There's 3 Other Stories You Probably Shouldn’t Ignore [VIDEO] - another great deep dive by AI Explained.

4. The Future Of AI Agents [VIDEO] - a funny and inspiring talk by HubSpot’s Dharmesh Shah about where AI agents are headed, with cool practical demos.

---

🔀 AI random

Other notable AI stories of the week:

1. A few big top-level OpenAI developments:

   1. The company will restructure into a “for-profit benefit corporation,” with Sam Altman potentially gaining equity for the first time.

   2. CTO Mira Murati announced her resignation, making her the latest senior executive to leave the company in the past year.

   3. …so did Chief Research Officer Bob McGrew.

   4. ...and VP of Research Barret Zoph.

2. Snapchat is partnering with Google Cloud to power its My AI chatbot.

🤦‍♂️ AI fail of the week

Come on, Gollum! You know it’s poor taste to outshine the bride! (Final version here.)

Anything to share?

Sadly, Substack doesn’t allow free subscribers to comment on posts with paid sections, but I am always open to your feedback. You can message me here:

---

💰 Sunday Bonus #29: How to create a NotebookLM podcast without using sources

As I made abundantly clear last week, I’m a fan of NotebookLM’s new “Audio Overview” feature that turns your sources into natural-sounding podcasts.

But did you know you can use this feature to create a podcast about almost any topic…even if you don’t have underlying sources?

I didn’t, either.

Well, turns out you can.

Here’s how.