ZukeMX ױ

Yesterday we spotted a strange phenomenon during #Qubic prototype mining. What looked like a phase transition during #AI training might actually be a realignment. This means that instead of adding guardrails we may move in the opposite direction, i.e. remove neurons like we remove cancer cells.

All know-hows will be revealed after qubic-network:native price touches $0.01, before that it's available on #Qubic mainnet only.

We have identified interesting features of #Aigarth's #AI training algo: 1. It creates hypotheses, verifies them, drops wrong ones, returns to put aside ones; 2. It uses the same pattern for different phenomena if they have similarities, applies little modifications to the pattern if necessary; 3. It chains simple calculations into complex ones, rearranges them. All this looks similar to how hoomans reason.

Hey kaspa:native holders. Let's play a game, I'll start shorting your coin and you longing it? Hit retweet if you are in or do nothing if you chicken out.

$QUBIC price is indeed not following current "pump" now. Because some holders are selling trying to jump on memecoin train (like $LAPTOP, haha). We know how it will end, I did the same with NXT and IOTA: We'll get rid of most of those who would sell at 10x and teleport straight to 100x-1000x range.

Remember when myriad universes were created, each with a distinct set of physical constants, and only ours gave birth to the humankind?.. Of course, you don't. Anyway, we (in #Qubic) have modified our $BTC price predicting models and now they behave like those universes. Only few, successfully predicting the price, will survive. This approach improved the speed of training by two orders of magnitude. It also sped up verification, so now we can scale 100-fold.

In #Qubic we have discovered something which looks obvious after you know about it: If #AI training is going slow then in the end it will likely fail to generalize the problem at hand and will overfit, if manage to get to a good value of the fitness function at all. Fast-trainable #AI successfully constructs a well-generalizing model in almost all cases.