Google Rejoins The AI Race

Bots Need a Supervisor

Hello, hello! 

You’ve probably heard and seen everything about Sora, OpenAI’s new AI model for videos. I’ll just leave a couple of links and skip right ahead to what you’ve subscribed for: top 1% AI news relevant to marketers. 

Official Sora Announcement: LINK

Sora vs Midjourney: LINK

A professional animator’s breakdown on the mistakes Sora made: LINK

My thoughts? It’s nowhere near perfect, but in its current form, it’s a great tool to pitch video ideas to clients. Makes everyone’s life easier.

On to the main newsletter now!

Google’s Back On Its Feet

Google announced a new model, Gemini 1.5, which beats GPT4 on several parameters. It won’t be available for regular, commercial use yet, and only developers can sign up for early access, but it’s still interesting to note and understand the capabilities of this model.

In this video, it shows how Gemini 1.5 Pro can understand, reason about and identify curious details in the 402-page transcripts from Apollo 11’s mission to the moon. What stood out the most to me is that AI can now identify humour. Makes me wonder how long it will be before it can write good humour?

Not scary enough? Here’s a video of Gemini 1.5 picking out a very precise moment from a 44-minute silent Buster Keaton movie when given a brief description. It also shows the model recognising a simple line drawing and using that as a reference to point out the exact instance when a real-life object based on that drawing is shown in the movie. 

Why is it relevant?

This essentially means that publicly available AI products will replace interns in the short run. We talked about this in last week’s issue, when we broke down Benedict Evan’s interview with Contagious. (ICYMI)

On a more positive note, while there will be fewer of them, entry-level jobs will be more interesting then they’ve ever been. 

If you’re curious about the model, check out Google’s official communication here: LINK 

Bots Need a Supervisor 

Yet another B2C company, which is a household name, was caught letting the bots go wild without supervision. Instacart, not to be confused by Instamart, which is ALSO a grocery app,  has been using AI to generate recipes photos with some hilarious result. 

A hotgod which looks like tomato on the inside, and a Chicken Insal which shows 2 chickens joined at the shoulder. 

Picture Credits: Business Insider

It’s a super brief read, check it out here. LINK

Why is it relevant?

It’s mostly just a chuckle-worthy story, but I wanted to take the opportunity to reiterate that humans overseeing AI content need to be fully present to catch the mistakes. We’ll see that the more passive tasks will get automated, and one will need to bring their 100% every single day. 

It’s Getting Easier To Build AI Assistants

Picture Credits: Olivia Moore on X

I have covered somewhat complex workflows involved in building AI Agents in the past, but this one is sooooo easy, I had to dedicate an entire section to it.

As a Partner at a16z, Olivia Moore takes a lot meetings on the daily. So she built a bot that sends her an email 15 minutes before a meeting, covering:

  1. Who she’s meeting

  2. Where they work

  3. Topics to cover 

She built it with a Zapier ChatGPT app, and she has described the entire process on her Twitter. It’s so simple, it’s beautiful. 🤌 

Granted, it took her a few hours to build it because she’s not a programmer. But it’s still very impressive! 

Read it here: LINK

I’ll see you next week with more exciting news, titbits, etc. about AI. 

It’s been nearly a month of running this newsletter. I end up doing quite a bit of research to draft this every week, and it’s honestly a great incentive to stay updated myself. 

All the same, if there’s anything specific you’d like to read, just reply to this email. 

Reply

or to participate.