Member-only story
IA / LAM Billion Dollar idea
Those AI on a Box devices like the Rabbit R1 if the idea is to get first to generate a bunch of training data for the Large Action Model, I think they are doing it backwards.
Because by marketing those things like an assistant they are setting up expectations of functionality and reliability unlikely to be meet until they have more data to train the model, which is going to be hard to get if the first impression of the device is “Unreliably thing”.
But, what if they try a different approach..
Because of its aesthetic I have hear the Rabbit R1 be refer as a Tamagotchi (a virtual pet for those born this side of the Millennium), so, Why not take advantage of that and make it a toy?
Imagine a cheaper version of the rabbit that connects to an app with Bluetooth where the game consist on scanning things in the real world and teach the pet about things on the real world and apps and how to use them, in essence to gamify the teaching mode so the players generate the data needed to train the Action Model and once is properly train then use that model to create the proper assistant.
Of course the hard part will be make a game worth playing, but if both Pokemon Go and the Digimon Vital bracelet could gamify *walking* maybe a game where you teach a virtual pet stands a chance.
Originally published at http://hashblogeando.wordpress.com on May 1, 2024.