This is Meredith, she’s the opensource ai we created to try and help solve this problem. If you were to think about her as a tiny pokemon, we have to train her to be the best pokemon (ad buying) ever! This means giving her moves and abilities, health points, and sending her into battle, playing her own little video game.
Meredith! Use [buy ad]!
When someone buys an ad, they have a ton of choices… what demographics should it be shown to, what location, what time? In a video game you may clearly know to cast fireball, but what party members need to be present to cast it, where should you be aiming, is the timing right? AI is really good at making these kind of decisions, so good in fact, that an opensource AI backed by Elon Musk is beating top ranked players of a 10 player multiplayer game today.
Interestingly enough, they also found that as they chipped away the milestones to victory learning was improved. It was not better to say win the game by doing these things (like destroying towers or securing kills), it was simply “win the game” “here are your moves”. Where before, the programmers were focused on multi-touch success indicators, removing focusing on those actually resulted in more streamlined learning. It is not that the ai stopped caring at all about kills. It is that as we focused her on simply winning, not getting a certain threshold of kills, she was able to see that metric for what it was, simply a means to win the game.
If we pivot our marketers cap around and focus on our true goal, selling burritos, our focus should simply be that, “win the game, here are your moves”. “winning the game” in this instance would be securing a conversion, Meredith’s moves would be the ad buying, organic reach, posting, etc. Instead of focusing on kills, we focus squarely on the end goal of securing a conversion, and see page visits (kills?) as simply that, a means to win the game.
Credit: Source link
