Jump to content

Your waifu in a neural network


Segai

Recommended Posts

So I was browsing YouTube and happened to stumble onto a guy who fed a recurring neural network (LSTM type) some cute VN girl noises. The girl in question is Hinata from Pure Pure. Needless to say the results are interesting...

It's rather interesting to see just what a machine picks up from that, and damn creepy.

The amount of iteration on the left signifies how often the network has gone over the 10 minute sample data and is also a good measure of how much it has learned.

Also bonus points if you turn on CC and watch a machine try to understand what another machine is trying to say.

Edited by Segai
Link to comment
Share on other sites

You know I thought the Pig - Human hybrid was the creepiest scientific thing I would hear about this month, but Jesus. Who knew that a machine desperately trying to learn Japanese would sound...... so fucking haunting. 

You know the more I think about it, this has a whole Minerva's Den vibe about it.

Edited by Soulless Watcher
Link to comment
Share on other sites

I remember seeing this last year.

Quite incredible. It's hard to imagine how far we got. It's even more thrilling to understand my generation might actually see the first AI's and perhaps, androids that won't be any different from us. that said, unless we'll branch off more into bionics and genetic engineering beforehand. There's a vast potential of different possibilities ahead of us.

Link to comment
Share on other sites

On 28/1/2017 at 1:27 AM, Soulless Watcher said:

You know I thought the Pig - Human hybrid was the creepiest scientific thing I would hear about this month, but Jesus. Who knew that a machine desperately trying to learn Japanese would sound...... so fucking haunting. 

You know the more I think about it, this has a whole Minerva's Den vibe about it.

Well, I advise you to disbelieve the pig/human hybrid.

The future is... artificial waifus :o

Edit: post 666! \m/

Edited by Okarin
Link to comment
Share on other sites

This is great and all, but what I'd really like to see is someone trying to generate a VN scene using a neural network. The fascinating thing about VNs is that, once you crack open the file format, you've got all the resources available and easily computer-generable (aside from voice acting... and let's just set the actual topic of this thread aside for the moment and ignore that...). So you can build whatever you want, or have a computer build it for you, harnessing extensive training from the huge amount of data available in any sufficiently long VN.

Or, better yet, train it on a big corpus of VNs (with some munging, like correctly genericized names that you then convert into actual names; hell, VNs use so many tropes you can even bucketize your genericized personalities) and then generate slice-of-life scenes from it. You probably wouldn't even be able to tell the difference between them and ones written by people! (Okay, yeah, you could certainly tell the difference, but with a bit of work, who knows...?)

I, for one, welcome our neural network-generated fandiscs.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...