Neural Style Transfer

Next

Previous 

Details


Ganesh H Shankar
Neural Style Transfer
More about this image in this interesting thread!
Thu May 21, 2020 8:48 pm
332
Ganesh H Shankar's CNP Gallery       |      Send PM to Ganesh H Shankar     |       [NEW] Recent Comments by Ganesh H Shankar

--
Ganesh H. Shankar
Wishing you best light,

Image
Personal Websites Fine Art Nature Photography
Facebook Pages Ganesh H. Shankar | Fine Art Nature Photography | Art Of Life


Ganesh H Shankar  Joined CNP On 24 Apr 2008    Total Image posts 682    -   Total Image Comments 6000    -   Image Post to Comment Ratio 1:9    -   Image Comment Density 39     -     Total Forum Posts 921

Rating & SHARING


not rated
Login to rate this image

Post a comment


Comments

Commentby dinesh.ramarao on Sat May 23, 2020 9:22 am

Ganesh,
I'm yet to understand deep learning, shall come back on this image sometime next week.
On the first look, can AI deliver creativity ? Art and Art'ificial go together?
-RD

Commentby Ganesh H Shankar on Sat May 23, 2020 12:31 pm

RD, sure, we will discuss this further. We need to explore what “human learning” really is and contrast that with deep learning. We also need to understand the role of “learning” in “new”/“creative” work. For now let me share a TED talk and a concert, of course authored by AI/deep learning.

I do think every life form in Nature has an ability to do some thing “new” without being trained. It can smile/be happy/sad without being trained. We all are born with differently coded genes which are not learned. This definitely adds to newer untrained part of the works. That is a wonderful thing which human creations can’t mimic. That said, if we just talk about “creativity” as something which appears like “pleasing” and incrementally “new” and has the baggage of “learning” from others‘ prior works or learning from schools/teachers then the boundaries are melting slowly! If you look at the generated image above carefully, no amount of tweaking the “photo merge” in photoshop can result in the newly generated pattern. There is no trace of tree left. The pattern however is influenced/learned/inspired by the curves of the branches and it appears “new” and unguided. One might argue that there is no relatable “human emotion” in these process of “new” creations. That would however also leaves us in the domain of individual taste and subjectivity. If 5 out of 10 novice listeners can’t make out whether these music shared in these links below are computer created then we may need to rethink about the notion of “creativity”. Probably the video/music below reinforce that better.




--
Ganesh H. Shankar
Wishing you best light,

Image
Personal Websites Fine Art Nature Photography
Facebook Pages Ganesh H. Shankar | Fine Art Nature Photography | Art Of Life



» Last edited by Ganesh H Shankar on Sun May 24, 2020 8:47 am; edited 10 times in total

Commentby dinesh.ramarao on Fri May 29, 2020 10:17 am

To the best of my understanding and opinion:

Human learning is not just data pattern matching, but to "intelligently" process them to "understand" and "respond" , there is an element of "emotion",

Now, so far, as on this day, Artificial Intelligence is pattern match and make decisions on the match and extrapolate "intelligence" to make it appear intelligent. The response is an action by AI, but to "respond" with "emotion" and "understanding" is still far away. Eventually, this could happen, we would have a emotionally wired robot.

Coming to this image, this is not what we both had in our mind, few years back when we discussed the "blend". Surely not even close enough, to my mind. The algorithms by Harish Narayanan are his intelligence ? or is it an intelligence factor learnt by his lending of "match" , I'm lost thinking in this line. While being in developing algorithms world for over two decades has given some intelligence to me to understand new algorithms, I failed to understand this Deep Learning, MY INTELLIGENCE failure? So be it, but, but, this is a step in getting heterogeneous "styles" blended with content ?

I think, these are our deep learning on new ways of creative work, but, can we give away creative thinking to technology?

I have questions and questions, have not found answers, will keep trying.

-RD

Commentby Ganesh H Shankar on Fri May 29, 2020 9:43 pm

The response is an action by AI, but to "respond" with "emotion" and "understanding" is still far away. Eventually, this could happen, we would have a emotionally wired robot.


RD, Coming to the emotions, will human creations (I mean robots/software created by humans) can have "emotions" in the sense that we smile, weep, feel sad, excited, frustrated....? I have zero hopes on this. We have not understood emotion well though we express them. I don't think that will ever happen. Just my guess..

Coming to this image, this is not what we both had in our mind, few years back when we discussed the "blend". Surely not even close enough, to my mind.


Oh, I just assumed then :) However, this is what I had in mind. I thought bringing together a frame with "clutter" and one with a few elements would be an interesting experiment. May be you had a different idea then?

I was focussing on the notion of "creativity" in this post.

Back to the definition of creativity,
creativity
noun
the ability to produce or use original and unusual ideas.

The point I wanted to make was, this ability of late is not restricted to human creations alone. If we have a 10 sound tracks, 5 of which created by a human artists and 5 generated by, say, AIVA, can one separate the two sets? After listening to some of these tracks I am not confident of someone succeeding using "lack of emotions" as a criteria to separate them! One may argue that AIVA used thousands of human creations to "learn". Similarly art created by humans too are influenced and "learned" over life of an artist. That "blend"/"mix"/"match" may become far more subtle. What probably will vary is degree of traceability to the original inspiration. I guess this chain of influences goes back to observing things in nature using our senses. Some may have receptors for some of this external stimuli and some may not. Those who have may "create" something appearing "new". However, the source/influence is still remains "external" and not "original" in some sense? So, I find it difficult to separate "human" learning from "deep learning". BTW, "deep learning" is not that nested "if then else" AI that we had learnt 30 years ago.

One day I just asked Amazon's Alexa -

me: "Alexa, when is my wife's birthday?"
Alexa: "You will be in trouble if you don't know that"

me:”Alexa, how do you work”?
Alexa: “Lots of people have lots to teach me, I am still learning more”

For the first question, I was expecting the answer "I don't know" from Alexa. I can't think of a programmer coding "list of all possible jokes/questions" to a device like this. Does it understand "humor"? What do we mean by "understanding"? Further, these systems keep “learning” (whatever it means) and answers over time for the same question may vary.

I think, these are our deep learning on new ways of creative work, but, can we give away creative thinking to technology?


I don't think the question of "giving away" arises here. Probably these innovations make us revisit our current definition of "creativity", "understanding", "learning" etc. We still have far superior "receptors" coded into our genes by Nature that transform an external stimuli in unknown ways. This I guess explains why we did not see million Bendres. That said, the influences behind human creations are undeniable. In that sense there is no absolute human "creation". I think Nature is the ultimate source of all creations. Our attempts at survival, which is what Nature wanted us to do, has led to our morals, ethics, good, bad and “art”. That I guess brings philosophy into picture.



For those who don’t understand Kannada, sorry, this short clip does not have subtitles. This is by Girish Karnad on the life and influences of one of our great poets, Da. Ra. Bendre.

--
Ganesh H. Shankar
Wishing you best light,

Image
Personal Websites Fine Art Nature Photography
Facebook Pages Ganesh H. Shankar | Fine Art Nature Photography | Art Of Life



» Last edited by Ganesh H Shankar on Sat May 30, 2020 12:10 pm; edited 7 times in total