IDEA #1 | VENTURE STARTUP

AI Video Detection Platform 📹

Finding the fakes

1. Problem/Opportunity❓

AI is moving at a blistering pace. Take AI video generation.

Back in early 2023 cutting edge AI text-to-video models were, to be polite, not so cutting edge. Check out this truly haunting video an AI model generated of Will Smith eating spaghetti.

But last week this changed, because OpenAI have thrown their hat into the ring. They launched their text-to-video model Sora, showcasing the model’s astonishing results.

There’s a problem though.

As AI generated video becomes more lifelike, platforms are going to need to be able to detect whether videos are AI generated or not for a wide variety of reasons, including content moderation. They will need to create their own models that have been trained on how to detect real or fake videos, ideally by humans, but doing this at scale is a massive challenge.

How are we going to do this? We’re going to take inspiration from something that’s been on the internet since the early 2000s…CAPTCHAs.

Let’s delve into the opportunity.

2. Solution ✅

Here’s the idea…create a platform which collects data about whether videos are AI generated or not which can be used to train models to detect fake, AI generated videos.

How will this work?

Well today when you’re solving a CAPTCHA (Completely Automated Public Turing test for telling Computers and Humans Apart) you’re not just proving you’re human, you’re actually labelling an image or a piece of text for Google. Ever wonder why CAPTCHAs are usually of objects on roads? You’re helping google’s self driving car models to understand what traffic lights and bikes are.

Let’s do the same thing for videos.

Here’s how the platform works. A company signs up to the platform and uses it to authenticate whether users signing in to their website are real or not, just like with CAPTCHAs. However instead of the usual CAPTCHA short videos are served to the user and they must judge whether the videos are AI generated or not

This data, combined with mouse movements (which are speculated to be used in reCAPTCHAs), would authenticate that the user is a human and also capture whether the videos served to the users are perceived to be real or not by humans.

This would be incredibly valuable training data which could then be sold to companies, like social media platforms, looking to build their own models to detect AI generated videos.

3. Business Model 🏦

Go-to-market: we think the go-to-market here is to build a website where people try to guess whether videos are fake or not in a quiz format to test the quality of their judgements and if successful proceed with a fundraise and full build

Monetisation: charge businesses for access to the training data

Startup Costs: costs for the MVP described above would be minimal. Once validated you’d need to do a raise to get this to a meaningful scale.

4. How You’ll Get Rich 💰

Scale this to a large business and sell to a large AI training data business like Scale AI

SPONSORED CONTENT (WE GOTTA PAY OUR BILLS)

IDEA #2 | STARTUP

Business Kit for Kids 🧒

Entrepreneurship…it’s child’s play

1. Problem/Opportunity❓

16:34.

2 days before the night before Christmas.

As usual I was perusing Twitter (I refuse to call it X come at me Elon) for my usual fix of unhinged opinions, devastating ratios and hilarious memes, when I came across a very interesting tweet.

Ask and you shall receive.

Lots of parents who are entrepreneurs themselves want to teach their kids about business but it’s something you can’t really involve your kids in, unless your last name happens to be Trump that is.

So let’s create a business which offers this to the rest of the parents out there.

2. Solution ✅

Here’s the idea…create a platform which enables parents to easily spin up a small side business which they can run with their kids in a controlled environment.

Here’s how it works. Parents sign up for the platform and decide what they want to call their business and all become “Directors” of the “company”. They can then design a logo on the platform using simple tooling.

They then decide with their kids what product they want to sell out of a selection of a few items through the platform. Once they’ve decided this then they can use a simple drag and drop interface to create their website and list their products for sale.

A small number of these products, along with packaging and shipping labels, are shipped to their house and then they can send their website live.

Once the website goes live it can be shared with friends and family who can then place orders and of course the parents can place orders from fake customers too which don’t get shipped.

You could also incorporate an AI accountant or taxman depending on how true to life you want the business to be, but let’s not have the IRS auditing a child’s e-commerce business, although i’m sure they’d relish in taking money from children if they could.

This could be a great bonding experience for parents and their kids while learning important business lessons along the way.

3. Business Model 🏦

Go-to-market: search twitter and find this guy, he’ll be your first customer. After that find business people on twitter who tweet about their kids, that’s your customer base right there

Monetisation: Parents pay a subscription fee as long as their “business” is active

Startup Costs: pretty minimal

4. How You’ll Get Rich 💰

This may not be a venture scale idea but this could be a great smaller startup or cash flow business. You may even be able to get a small exit here.

NO BRAINER IDEA

Alternative Authentication

Face ID and finger print authentication have been in phones for years now. But they don’t always work very well, like when we’re wearing gloves of a face mask.

Introducing a brand new way to unlock your phone…by licking it.

All you do is lick your screen and it recognises your unique tongue print so you can unlock your phone. Efficient, hygienic, what more could you want.

BIG DECK ENERGY

Bolt’s $393m Series D Deck

Year: 2021

Stage: Series D

Amount: $393m

Bolt offers one-click checkout software that could plug into the existing payments platform of any business. In 2021 they raised $393m at a $4bn valuation, however since then their valuation has fallen 97% from its all-time high of $11bn. Checkout (pun not intended) their deck by clicking on the link below.

JUST THE TIP

The Twitches are in the Niches

Twitch is a pretty unique platform.

It’s kinda like a country with its own culture, history and even it’s own language with users throwing around terms like emote, poggers and meta that we’re too out of touch to understand.

Twitch was founded back in 2007 by Justin Kan and Emmett Shear, but the platform looked very different back then. It was originally called justin.tv where, Justin attached a webcam to his head and streamed live to the internet for eight months straight. This was frankly, a terrible idea and the team pivoted to allow anyone to stream on the platform.

It worked…kinda.

For a few years the platform grew but then growth started to stagnate and the company ran into legal issues since many users streamed sports events in its early days. Sports leagues aren’t big on piracy.

The team realised they needed to focus their efforts on a single niche. As the saying goes, “the riches are in the niches.”

After much internal debate and despite Kan’s hesitance, the team decided to focus on the gaming niche and in 2011 spun-off twitch.tv as a live streaming platform for gaming and the platform was acquired by Amazon in 2014 for $970bn. Since then the company has grown massively and re-expanded back out to other verticals beyond gaming.

You can read a detailed account of the early years of Twitch here, a fascinating tale with police raids, piracy and drugs all involved.

But is there a lesson here?