One App or One AI

Over the next 5 years, AI will drastically change how we interact with our devices. Eventually we will be able to perform any and every task from one “Homebase”. The question is how? And will that make be any different?

The later is a fair question, because our phones already are the center of our universes. It’s where you can book and Uber, buy flights, order food delivery, check your email, and so much more. But in 5 years, a phone without a Homebase will be as foreign as a phone without a touchscreen.

My hypothesis is this Homebase will exist in one of two forms. Either one app to rule them all, a super app. Or one almost AGI quality AI to rule them all, similar to the movie Her (which disclaimer, I have not watched).

The case for a Super App

Lot’s of countries already have Super Apps, i.e. WeChat, Gojek, Grab. WeChat is everything in China. It’s used for setting up a doctor’s appointment, managing bank accounts, ordering in person at restaurants, and literally everything else.

What I want to talk about here is what it will take for a super app to rule the US. This hypothetical company (which I will call Aster) must:

1. Start with money - In the US we put brands in a box fast. Aster’s can’t go from social to money like we chat because it is too large of brand shift. There is an inherent distrust with social media. So Aster must begin with an area that demands trust, like Banking.

2. Form Hidden partnerships with existing brands - The VC fueled gig economy is perfect for Aster. It means there are tons of large, unprofitable, companies that need to get distribution by any means (i.e. Instacart and Uber Eats recent partnership). Aster needs to deploy integrations while only highlighting the Aster brand. These must be marketed as Aster can now deliver food, not Aster + Doordash can deliver food.

3. Enable Account Aggregation - Aster’s brand comes above all. Therefore, in order to power integrations, Aster must support either aggregating existing accounts or creating new accounts. So basically as 2020’s version of sign in with Facebook that happens in one smooth motion.

The case for baby AGI

My favorite ways to look at AI is like 100,000 tiny humans. With this lens one AI must win. If I 10 personal assistants in a group chat to offload my tasks too, life would be easy.

This hypothetical baby AGI (which I will call Her) must:

1. Search like a human - Basically what Rabbit AI was claiming to be able to do. The frame of reference for Her hundreds of tiny humans, so Her must interact with websites built for humans like a human would. MultiOn is a cool example of someone combing AI and playwright to achieve a similar effect.

2. Highlight the brands - AI must recognize brands like Kayak, Uber, Amazon, etc. Conversely to Aster, Her’s goal is to build trust in it by leveraging existing brands. A lot of work needs to go into making sure the companies Her accesses get recognized by the users executing the command.

3. Activation Anywhere - Her needs to be omnipresent. Activation anywhere by any method (check out how Apple imagines this level of activation). Her must integrate with the default voice assistants on phones, integrate with command bar on mac, in chat on apps like IG or Tiktok. But even more importantly, it needs to be able see what you are doing at all times on every device. That way it can queried naturally as if you had tiny humans ready peering over your phone and desktop at all times.

Aster builds tech that bends other companies to its will. Her builds AI that bends AI to humans will.

I think it’s easier to persuade tech than people.