They are the most creative non-biological systems in the world. From deepfakes to musical scores from building architectures to new faces can be made with GANs. GAN stands for generative adversarial network. They come under RNN’s(Recurrent neural nets). GAN is a framework put forward by IAN Goodfellow and his team in 2014 and it has established itself as a subfield itself. Gans are some of the most fascinating pieces of software code I had ever seen.

They have a really peculiar structure in which two neural nets compete in a zero-sum game. The generator produces new results and the discriminator distinguishes between them and if it finds the result non-convincing concerning the real samples the results are rejected. The main goal is to produce convincing results which are close to training data and as they compete better and better results are found. Input can be changed to get incredible results. New artists are emerging who use gans to produce art which you probably will admire and may even think was made by some great artist. GAN’s are being widely used in many fields- from modeling the dark matter distribution to producing lasting structures. But the application of this tech that excites me the most is their ability to generate structures and models of things like cars, bridges, and buildings. These models won’t be built by labor but by 3d printers. Some structures have already been built which were half the weight and lower price than their old counterparts(human-made and designed). We can ask these neural nets to produce a car chassis half the weight and more strong and it might help us build a safer future. here is a link to the colab notebook where you can run code and listen to the music generated by gan itself-

https://colab.research.google.com/notebooks/magenta/gansynth/gansynth_demo.ipynb#scrollTo=rSgo0YiVUm17

I was unable to add an audio file to the newsletter So just shared the code.

Take a look at this picture and you will think why is it being shown in a blog post relating to AI. To your confusion ladies and gentlemen, it’s created by a Neural net. You might try hard now to find errors and good luck with that. No person to the record(as I read) has a face like this. This does draw a lot of criticism for future wars being caused due to deepfakes of politicians and fake ids. And once again its use and effect depend on how we use it. There is a great ted talk relating to this subject -

I really liked it and hope you did too. After the video and reading the blogpost I urge the reader to just put 1 minute into the thought of how GAN’s could help you.

PS: Sorry for posting it late. You can show me a gesture of appreciation by buying me a book so I can read and write more cool blogposts .

https://www.buymeacoffee.com/madhavgoyal

They are the most creative non-biological systems in the world. From deepfakes to musical scores from building architectures to new faces can be made with GANs. GAN stands for generative adversarial network. They come under RNN’s(Recurrent neural nets). GAN is a framework put forward by IAN Goodfellow and his team in 2014 and it has established itself as a subfield itself. Gans are some of the most fascinating pieces of software code I had ever seen.

They have a really peculiar structure in which two neural nets compete in a zero-sum game. The generator produces new results and the discriminator distinguishes between them and if it finds the result non-convincing concerning the real samples the results are rejected. The main goal is to produce convincing results which are close to training data and as they compete better and better results are found. Input can be changed to get incredible results. New artists are emerging who use gans to produce art which you probably will admire and may even think was made by some great artist. GAN’s are being widely used in many fields- from modeling the dark matter distribution to producing lasting structures. But the application of this tech that excites me the most is their ability to generate structures and models of things like cars, bridges, and buildings. These models won’t be built by labor but by 3d printers. Some structures have already been built which were half the weight and lower price than their old counterparts(human-made and designed). We can ask these neural nets to produce a car chassis half the weight and more strong and it might help us build a safer future. here is a link to the colab notebook where you can run code and listen to the music generated by gan itself-

https://colab.research.google.com/notebooks/magenta/gansynth/gansynth_demo.ipynb#scrollTo=rSgo0YiVUm17

I was unable to add an audio file to the newsletter So just shared the code.

Take a look at this picture and you will think why is it being shown in a blog post relating to AI. To your confusion ladies and gentlemen, it’s created by a Neural net. You might try hard now to find errors and good luck with that. No person to the record(as I read) has a face like this. This does draw a lot of criticism for future wars being caused due to deepfakes of politicians and fake ids. And once again its use and effect depend on how we use it. There is a great ted talk relating to this subject -

I really liked it and hope you did too. After the video and reading the blogpost I urge the reader to just put 1 minute into the thought of how GAN’s could help you.

PS: Sorry for posting it late. You can show me a gesture of appreciation by buying me a book so I can read and write more cool blogposts .

https://www.buymeacoffee.com/madhavgoyal