😺 Video-To-Text

PLUS: Google screenshots LEAKED

Noah Edelman & Pete Huang
October 27, 2023

Welcome, humans.

It’s Halloweekend, which means lots of bewildering Halloween costumes that’ll leave you scratching your head like a “sexy” Oppenheimer-themed Barbie. Your safest bet? Describe the costume to ChatGPT and let it dish out some educated guesses on what it might be.

Here’s what you need to know about AI today:

A cutting-edge tool that understands videos was unleashed.
Google is set to make a huge AI push in 2024.
ChatGPT’s cutoff might’ve been moved up to April 2023.
YouAI created a tool that transforms books into chatbots.

AI That “Watches” Videos Is Here.

Current AI models understand text, PDFs, images, and spreadsheets.

This week, Twelve Labs launched “Pegasus-1”, a model that understands videos!

(here’s the technical report for ya nerds).

Translation: Pegasus can “watch” a video and tell you what happened.

For instance, you could upload a Mr. Beast video and quiz the model on how facial expressions and props contribute to the content. Or you could have the AI recap any YouTube, TikTok, or camera footage for you.

Why it matters: This creates an entirely new category of applications, cause, until now, robots were basically blind. Imagine the applications this could unlock:

Content Moderation: Automatically flag questionable videos based on elements in the footage (e.g., firearms).
Product tagging: automatically tag products in promo videos.
Work Safety: Keep tabs on safety compliance in industrial settings like factories.

Just imagine security cams juiced up with this tech — swiping a single sock from the laundromat just got riskier!

Super stoked to see what comes out of this. Get yourself on the Pegasus-1 API waitlist, stat!

FROM OUR PARTNERS

Here’s An Easy Way To Reach 4x More People (hint: dubbing).

Only 20% of the world speaks English as a first language, meaning your business is literally missing out on billions of potential customers.

ElevenLabs’ new AI dubbing feature can convert spoken content to other languages in minutes, all while keeping your original voice intact.

Game. Changer.

This means all of your…

YouTube/TikToks.
Marketing campaigns.
Podcasts.
Welcome videos.

…can now be translated and tailored for a global audience with a few clicks.

This tool is amazing. We highly recommend you test it out yourself — stat!

Google Is Betting Big That 2024 Is Its AI Year.

It’s been a tough week for Google investors, with $GOOGL down nearly 10%.

*collective sigh as everyone opens their eTrade accounts.*

It’s in part cause Google lacked the numbers to back up its AI push, unlike its foe $MSFT, with its Azure OpenAI service already being used by 18K companies.

Just wait until 2024, Google says…

Much hinges on the success of Google's ChatGPT rival, "Gemini". Fresh leaks confirm that Gemini is poised to replace PaLM, the engine behind Bard.

Gemini will be multimodal, like ChatGPT-4v, but jury's still out on how well it stacks up.

The leaks also revealed "Stubbs," a new AI tool that can build entire apps, design and code included! No code, just prompts. Neat!

Google’s been playing it slow and steady (they originally said Gemini would be out this year). Maybe it all pays off with Gemini and Stubbs?

Around the Horn.

OpenAI is giving away $25K in API credits for the top 10 submissions aimed at promoting AI preparedness.
ChatGPT’s knowledge cutoff is potentially now April 2023.
How one company is using GPT-4 to streamline user surveys.
YouAI created an app that turns books into AI chatbots that you can talk to (here’s how).
Our sponsor today: ElevenLabs. Our first story today: Twelve Labs. Fun fact: they hosted a hackathon together called 23Labs.