Happy Monday, everyone.

As always, we love your feedback on how we can improve.

Don’t hesitate to reach out.

🤖 Top Stories

🚀 Pinecone AI raises at a $700m valuation

Imagine if you could group everything on earth by similarity.

Not just objects - pieces of work, ideas, tweets, concepts.

The dimensions you’d need to store that type of data would be astronomical. Thousands of different dimensions. It’d be impossible to tag them all, much less store it.

Pinecone, an AI startup, aims to store that data. They’re the buzziest startup in a new AI category known as vector databases.

How big is this space? Their latest funding round valued them at $700 million.

Let’s back up. Vector databases store and manage high-dimensional, unstructured data. These vectors represent complex relationships or patterns.

Humans are bad at analyzing these.

AIs are good at analyzing them, and finding clusters and similarities in the complex data.

Imagine finding every sentence that sounds like this one. Difficult. But vector databases can store and query this information.

Bottom line: AI will be more powerful. Vector databases will fuel their ability to do that.

👀 ChatGPT app when?

The long-awaited ChatGPT app might be closer than we think.

Two job postings popped up on OpenAI’s website, both on the ChatGPT team.

“Software Engineer, iOS” and “Software Engineer, Android”.

The job descriptions are not interesting. What’s interesting is that OpenAI is finally doing it.

ChatGPT is the perfect candidate for a mobile companion app.

For all the hype that ChatGPT has, it’s still niche. Only 17% of Stanford students used it to assist in their fall exams. Most people haven’t used it at all.

A mobile app might be what OpenAI needs to get it in the pocket of every techie.

And if you apply, tell ‘em the robots sent you, okay?

🤯 ChatGPT is the remote for your AI requests

Researchers from Microsoft research in Asia had a novel thought. ChatGPT can serve as your LLM concierge.

What does that mean?

Right now, ChatGPT can only use…itself. GPT-4. Nothing else.

Instead, when you send a prompt to ChatGPT, it will

1. Interpret the request

2. Plan out the execution

3. Decide which model(s) are best for the job

4. Handle orchestration of the tasks

5. Summarize the result and send it back to you

Their example shows how they sent ChatGPT + HuggingFace a photo. They wanted ChatGPT to tell them what is in a picture, and how many of each object there are.

ChatGPT used a few models from Google and Facebook to get there, and it worked.

OpenAI might be trying to create an everything app with ChatGPT. Papers like this show it’s not that far off.

😎 AI tools, tips, links

1. Use GPT-4 to make a video game. Learn more

2. ChatGPT plugin examples. Learn more

3. Use ChatGPT to explain code. Learn more

4. Execute code with Auto-GPT. Learn more

5. 1000+ GPT-4 prompts. Learn more

(AI) family is more important than anything.

🤖 Got feedback for us?

Head over to our Twitter, @RobotRockDaily, and let us know what you’d like to see more or less of.