You are viewing a single comment's thread.

view the rest of the comments →

20437377? ago

I've felt like I innately just have a gut feeling that this is true, or at least very close to the truth.

I don't know why, except to say that I do have somewhat of a decent idea of how far AI has come (we use it in digital marketing like crazy anymore in a variety of ways) and I have to assume that the government (who invented the internet to begin with) uses the most advanced version(s) of it for all sorts of shit, not the least of which being something so big as waging war/controlling people.

Just curious, where (if anywhere at all) are you getting this info, OP?

20438784? ago

Keep in mind the "government" is not separate from FB/Google, these things were born out of the CIA with taxpayer funds. The deep state is, or has been until now, the de-facto "government". The Old Guard.

Q is the spiritual forces at the next level above this. Angels. "We have everything."

I don't think it's AI vs AI.

It's AI vs II (infinite intelligence). II = God. God wins.

20445830? ago

Pardon me while I semi-hijack this high-ranked post. Downvoat if you must, but as someone who knows more than a little something about this field, I'd like to point something out. The "adversarial" in GAN specifically refers to 2 competing networks only within the scope of training the discriminator network used to then predict how likely seen features are, given some assumption of the post (in this case). The generator network creates data and the discriminator network then evaluates the data and likelihood of it being in the training dataset. Think of it as network co-evolution. This adversarial training all happens in one scope (i.e. the deep state, google, facebook, whichever organization or group is building the network).

OP's post is still correct no matter what kind of training method used to build a network. You can simply substitute "ML network" or "AI network" for "GAN" and the info in the post still stands. Do not get hyper focused on GANs and do not mistake the "adversarial" name as being inherently bad.

Again, sorry for the hijack, but this needed to be seen and not lost at the bottom of ~150 posts.

20448199? ago

Shit in Shit out.

20450169? ago

100% true. Bad training data makes bad models.