Well, it is interesting in a “neurons that fire together, wire together”-kinda way, no doubt. I supposed it could be used in a brain storming kinda way or to raise awareness of your own selection bias… people should post here if they find a way to make could use of it. There seems to be some sort off potential… can’t put my finger on it though at the moment.
If you look at the instructions, for the “arcade” game… the word you type that is most associated gets sorted to the bottom. It’s sort of like a frequency dictionary in this way or what you’d find on the SATs with “a is to b is like c is tod”-questions.
So depending on what the other words are in the list, “Italy” might sort “pasta” down to the bottom, but if other key terms related to Italy are in the list “pizza” might to a better job, the opposite is true when there is a bunch of food items in the list of course.
The AI “knows” the association based on the training it’s received and maybe on the training it continuous to receive as people play it. Not sure about the latter, but then you could create a bias for the AI to think “yellow” and “pee” is more associated with “snow” than is “winter” or “ice” or “white” is.
So naturally after playing a couple of games I tried to see how often “a”, “the”, or “be” would sort a given word to the bottom. Then I got really bored (first encounter with AI programming 20 years ago) and tried “sex”. Well, try for yourself if you’re curious… by the way, here’s why:
You can try the “blocks game” which doesn’t have a timer. There’s actually two under that link… the first one that you’re talking about which is like Tetris… five right answers clears the board; and the second one (without timer) that is like candy crush. I think depending on what you are trying to do both have they benefits.