Your examples are pretty bad, you listed two activities that purely rely on statistical outcome and exact mathematics, ai art will always be behind humans, always, no matter how many time it passes, you clearly dont undestand how ai and machine learning works.
Recommending Breast Expansion Games Archive (2021)
- Thread starter hibilly
- Start date
Endgame AI art (and machine learning) is not imitating humans perfectly. It's collecting the best characteristics and combining them. No artist is simultaneously the best at everything. This is only with the assumption that it can't beat the images it's trained on, which is also wrong.
Probably a dedicated topic, as that would also add it to the games list as well as serve as a place for questions and bug reports.
Yall are talking machine learning. Figured i would chime in.
Actual deep learning for visuals can be done using a Generalized Adversarial Network (GAN). You can think of it as one part with training using supervised learning for the initial curriculum, but then an Adversarial Network introduces noise on another epoch. This second epoch is can be supervised, but the noise forces better and better learning. Many iterations of this is done. This is the type of system that can make deep fakes.
LLM actually use what is called a transformer neural network. It can do more advanced symbolic manipulation, like interpret natural language and analogies, but it is piping a VAST amount of information. The transformer is like a search engine wrapped in some kind of "canonical ensemble" piper. The later follows the rules of grammar, the prior can recombine piped information to derive more accessible unique information. It does not really "compress".
Though compression brings in other interesting points. Landauer princple (bit erasure), Kolmogorov(complexity), Shannon-entropy(new information), Kullback-Leibler divergence (which distribution), etc.
I mean images on a screen will always be a configuration space. Yes ML will bring more scalability, but the primitives of the reversibility of nondeterministic computing and irreversible (information is conserved or reduced) computing will remain. Noise computing actually can create, destroy, or conserve information. The nature of scalable information will remain interesting, though still on the basis of geometric axioms, which themselves can refute information entirely. This doesn't matter if it is ML or quantum or what have you, though those things could further scalability. We could sill have to remain humble before simple ideas earlier in our epistemology.
The word water is not wet. And yet it only takes to sticks to lean, to start a story.
Actual deep learning for visuals can be done using a Generalized Adversarial Network (GAN). You can think of it as one part with training using supervised learning for the initial curriculum, but then an Adversarial Network introduces noise on another epoch. This second epoch is can be supervised, but the noise forces better and better learning. Many iterations of this is done. This is the type of system that can make deep fakes.
LLM actually use what is called a transformer neural network. It can do more advanced symbolic manipulation, like interpret natural language and analogies, but it is piping a VAST amount of information. The transformer is like a search engine wrapped in some kind of "canonical ensemble" piper. The later follows the rules of grammar, the prior can recombine piped information to derive more accessible unique information. It does not really "compress".
Though compression brings in other interesting points. Landauer princple (bit erasure), Kolmogorov(complexity), Shannon-entropy(new information), Kullback-Leibler divergence (which distribution), etc.
I mean images on a screen will always be a configuration space. Yes ML will bring more scalability, but the primitives of the reversibility of nondeterministic computing and irreversible (information is conserved or reduced) computing will remain. Noise computing actually can create, destroy, or conserve information. The nature of scalable information will remain interesting, though still on the basis of geometric axioms, which themselves can refute information entirely. This doesn't matter if it is ML or quantum or what have you, though those things could further scalability. We could sill have to remain humble before simple ideas earlier in our epistemology.
The word water is not wet. And yet it only takes to sticks to lean, to start a story.
- Jul 23, 2022
- 9
- 13
I would argue that ai is already better than the average human. obviously it wont be better than actual artists but unlike artists its much more versatile being able to can change styles on the fly, and is completely free as long as you have the PC to run it. not to mention LoRA's are easy to make and can replicate anything thrown into them giving you basically an infinite library of characters, art styles, scenery etc. at a couple presses of a button.
Can we get back on topic? There is another thread about AI art.
I really like this game, this is definitly my kink. I really like how my actions transform day by day my character and somehow it makes me want to search through all the 12 000 images. I would say some things about the game:
- The only real way to enjoy the game for me was earning money with streaming sesions. All other ways are making me feel extremely poor aaaaall the time. Probably I would add some "extra life" so the landlord dont kick me right away the first day without payment. The same thing about breakfast. It is somehow hilarious that I get up in the morning and decide that my life has ended cause i run out of butter for my bread. Probably better some other kind of consecuence, maybe as you already do in other places when character is forced to modify her body as a punishment of some sort (streaming for example)
-As the game is about grinding the stats, I found the transitions between scenes a little bit slow, probably I would rather stay at the gym after finishing the sesion to start another one instead of going home and back to the gym. This happen with at a few other places.
- The Ai images are mostly very nice. When the top of all stats is reached, things get out of hand really hard. Especially with Ass Expansion. Is not just a hand fusing with a breast or that kind of thing, there are true madness over there. Aside from those images, the flow from different stats feels nice, maybe it make me feel weird when the caracter have suddenly a very childish face.
- Descritions at the mirror are a nice add to check your current level of each stat, but I must point out that weight gain is not feeling right. The comments are to harsh, especially when BE, AE, MG and Weight gain usually have a common consumer, so i would go for a more positive aproach or at least, neutral aproach. Maybe even more important to this point is the fact that until the last two levels more or less, the character is actually thin. It is kind of weird reading that im simply fat with a flat stomach
The rest of the ideas I have to tell is your WIP, llike other locations, interactions, even clothes. You have already throw hints about your line of development. As I said at the start, I really enjoy this game, it quite scarce the amount of content of any kind of expansion, yours is really wellcome!
Going to agree with this, especially image cleanup and consistency corrections.
For anyone having trouble with money/grinding/weight bad end, the way I go through the start is waitress, buy the cooking book and learn that while eating free waitress food, then learn the second cooking book while still eating free food, then start reading the book for the desired career path so I can start a job with a higher income to offset rent and occasional book buying. Once in the job with positive net income, I go for the third cook book and can start managing weight.
The skip dialogue setting and ctrl key are very useful for this stuff.
- Oct 16, 2020
- 573
- 1,169
i disagree.
compression is essentially what it is doing. the trained model (the actual file, such as the ones you can download on huggingface) is essentially a compression of the training data. because the file, despite how large it is, is still way way smaller than the training data while still giving output on par with the training data (with some margin of error).
it's basically lossy compression, just very different than traditional compression.
back to thread. does anyone here following the hypnosis dev for a long time know if he's slowed down on releases or if it's always been this slow? last one took almost a year if i remember right.
it's my fav BE game but holy fuck it never gets any update. their other game "endowed" which was also slowly introducing BE also seems to be in limbo.
Last edited:
Oh my gooood, go speak about AI in the right thread !
Can we please go back to posting games with growing titties?
- Jul 19, 2021
- 144
- 763
This released a while ago, and it's really short, but I don't think it has been posted here: No Nut November: A Futa Awakened. As the name suggests it's focused on Futa, but also has BE and GTS, and the dev doesn't joke around with any of those fetishes.
I think progress probably has tended to slow down, and of course having the dev's attention divided between two games isn't going to help. Finishing the Hypnosis Broken Watch branch took a long time, but I'm not sure that things have speeded up at all since. He seems to spend a lot of time on side activities like producing picture sets of the Hypnosis girls. He may have reached the stage where plot ideas are becoming harder to come by.
- Oct 16, 2020
- 573
- 1,169
seems it's that stage where devs are both not interested in dev work anymore but also feeling like they have to do it to sustain their income.
same thing with kingdom of deception and many other games. pig princess hurts because there's an anal-only path but we get a small update every 6 months.
Escape from the Room of the Serving Doll [Final] [MomoYama Productions] This game has the transformation tag, and it talks of a breast expanding robot but there's no BE in the game as far as I can tell, just huge breasts. Still a fun game, but a tad disappointed.
Their is a BE game called Too Lewd, would anyone take one for the team and buy the android edition plsss
You must be registered to see the links
- Oct 16, 2020
- 573
- 1,169
thanks.
played it for a while but only saw 1 BE scene. there should be at least a 2nd one but couldn't figure out how to unlock it before i gave up.