Neural networks and design in the future. Will artificial intelligence be able to replace the designer?
Hi, my name is Alex, I have been working with Photoshop for more than 8 years, and of course, I follow new technologies that are related to design. The article is by no means written to sell you a course. Initially, we wrote an article for the second design team, which has not yet worked with neural networks (as you understand, I have several teams, and we do not only make arts).
To begin with, neural networks now very rarely generate images that businesses need. 95% of commercial work at the moment the neural network is not able to create without the global intervention of the designer. Neural networks do not know how to make corrections to work, and at the same time, in the normal work of a designer without neural networks, 50-90% of orders give edits. Imagine how many attempts of image regeneration it will take to make corrections to the order. In my experience of working with neural networks for more than 3 months, I have not been able to generate a single work on the client's technical assignment in full. I made a lot of corrections to each image from the neural network. And I can't even imagine how much time and attempts it will take to move the house on the conditional image by 10 pixels, when the basic concept of the image as a whole has already been accepted by the client. Neural networks do not split the image into layers, they cannot adjust the degree of blurring, brightness and many other parameters that a person changes in Photoshop. And in the future, it is unlikely that something will change globally. And if there is a breakthrough technology, then someone has to work with it anyway, and that person will be a designer.
A designer is needed because most clients do not give clear technical tasks, and the designer has to analyze the client's product/website and create a design based on these visual data. Neural networks do not work completely like the human brain, they cope very poorly with visual data. They can't analyze individual design elements that need to be emphasized.
Now I will show with my example how I wanted to create a similar art in mijorney. The example below is the art I created in Photoshop without using neural networks:
To begin with, neural networks now very rarely generate images that businesses need. 95% of commercial work at the moment the neural network is not able to create without the global intervention of the designer. Neural networks do not know how to make corrections to work, and at the same time, in the normal work of a designer without neural networks, 50-90% of orders give edits. Imagine how many attempts of image regeneration it will take to make corrections to the order. In my experience of working with neural networks for more than 3 months, I have not been able to generate a single work on the client's technical assignment in full. I made a lot of corrections to each image from the neural network. And I can't even imagine how much time and attempts it will take to move the house on the conditional image by 10 pixels, when the basic concept of the image as a whole has already been accepted by the client. Neural networks do not split the image into layers, they cannot adjust the degree of blurring, brightness and many other parameters that a person changes in Photoshop. And in the future, it is unlikely that something will change globally. And if there is a breakthrough technology, then someone has to work with it anyway, and that person will be a designer.
A designer is needed because most clients do not give clear technical tasks, and the designer has to analyze the client's product/website and create a design based on these visual data. Neural networks do not work completely like the human brain, they cope very poorly with visual data. They can't analyze individual design elements that need to be emphasized.
Now I will show with my example how I wanted to create a similar art in mijorney. The example below is the art I created in Photoshop without using neural networks:
Neural networks and design in the future. Will artificial intelligence be able to replace the designer?
Neural networks can't come up with something creative (at the same time meaningful) based on the client's product, because the neural network learns from the models that are loaded into it initially. And from a legal point of view, neural networks are trained on images of other designers taken from the Internet, and then the question of copyright arises. In the nearest future, the creators of neural networks can be sued and it is possible to close or ruin some projects, or slow down their development, because there will be little investment due to a large percentage of money deductions to designers, not investors. This is just my reasoning. Time will show.
Article update: Getty Images has filed a second lawsuit against the developer of Stable Diffusion for using a photobank to "train" a neural network — this time in the USA. According to the photobank, the company copied more than 12 million of its images, thereby infringing on copyright.
Now let's analyze neural networks as a tool for a client who needs a banner or collage for a website.
Let me remind you that the client pays money to the designer so that the designer spends his time and offers ideas on how to make the design better. But let's imagine that the client paid money for access to the neural network, instead of the designer's services. When a client who does not work with neural networks tries to generate images, he will spend a lot of time writing queries. The big question remains, but how will the client be able to upload photos (photos of additional objects) to the neural network, now neural networks known to me do not have such functions. But let's say the client somehow uploaded photos, wrote a technical task for a neural network... and with a high probability the client will not get anything good. Requests need to be able to compose, describe every little thing, and this is a lot of time. In addition, you need to have taste, you need to study design trends in order to write a good query for a neural network.
Will neural networks be able to help the designer?
Of course, neural networks can help a designer generate some individual objects, a background, but you need to be able to edit all this, create a composition. Therefore, I propose to take the course «Commercial ART Photoshop course from 0 to PRO», where you will learn the basic method of image editing, working with collages, and learn how to create beautiful art from a variety of elements. You will learn up to 90% of the most necessary functions in Photoshop. Follow the link and acquire advanced knowledge!
Article update: Getty Images has filed a second lawsuit against the developer of Stable Diffusion for using a photobank to "train" a neural network — this time in the USA. According to the photobank, the company copied more than 12 million of its images, thereby infringing on copyright.
Now let's analyze neural networks as a tool for a client who needs a banner or collage for a website.
Let me remind you that the client pays money to the designer so that the designer spends his time and offers ideas on how to make the design better. But let's imagine that the client paid money for access to the neural network, instead of the designer's services. When a client who does not work with neural networks tries to generate images, he will spend a lot of time writing queries. The big question remains, but how will the client be able to upload photos (photos of additional objects) to the neural network, now neural networks known to me do not have such functions. But let's say the client somehow uploaded photos, wrote a technical task for a neural network... and with a high probability the client will not get anything good. Requests need to be able to compose, describe every little thing, and this is a lot of time. In addition, you need to have taste, you need to study design trends in order to write a good query for a neural network.
Will neural networks be able to help the designer?
Of course, neural networks can help a designer generate some individual objects, a background, but you need to be able to edit all this, create a composition. Therefore, I propose to take the course «Commercial ART Photoshop course from 0 to PRO», where you will learn the basic method of image editing, working with collages, and learn how to create beautiful art from a variety of elements. You will learn up to 90% of the most necessary functions in Photoshop. Follow the link and acquire advanced knowledge!
I gave different descriptions for the operation of the neural network, before that I fully familiarized myself with the instructions for the bot and videos - lifehacks. Here's what happened as a result:
The result was terrible, while I changed a lot of descriptions and spent about 2 hours on these experiments. You yourself can experiment with my photo and add a description: art with palm trees, I'm sitting on the sand, sunny weather. Even with any other description, the result is terrible.
Below I have added the results by description: ART - London 1930. The face in general is very much distorted, as in the previous results. Such work would never be accepted by customers.
Below I have added the results by description: ART - London 1930. The face in general is very much distorted, as in the previous results. Such work would never be accepted by customers.
Here is my photo, which I uploaded to the neural network:
Hi, my name is Alex, I have been working with Photoshop for more than 8 years, and of course, I follow new technologies that are related to design. The article is by no means written to sell you a course. Initially, we wrote an article for the second design team, which has not yet worked with neural networks (as you understand, I have several teams, and we do not only make arts).
To begin with, neural networks now very rarely generate images that businesses need. 95% of commercial work at the moment the neural network is not able to create without the global intervention of the designer. Neural networks do not know how to make corrections to work, and at the same time, in the normal work of a designer without neural networks, 50-90% of orders give edits. Imagine how many attempts of image regeneration it will take to make corrections to the order. In my experience of working with neural networks for more than 3 months, I have not been able to generate a single work on the client's technical assignment in full. I made a lot of corrections to each image from the neural network. And I can't even imagine how much time and attempts it will take to move the house on the conditional image by 10 pixels, when the basic concept of the image as a whole has already been accepted by the client. Neural networks do not split the image into layers, they cannot adjust the degree of blurring, brightness and many other parameters that a person changes in Photoshop. And in the future, it is unlikely that something will change globally. And if there is a breakthrough technology, then someone has to work with it anyway, and that person will be a designer.
A designer is needed because most clients do not give clear technical tasks, and the designer has to analyze the client's product/website and create a design based on these visual data. Neural networks do not work completely like the human brain, they cope very poorly with visual data. They can't analyze individual design elements that need to be emphasized.
Now I will show with my example how I wanted to create a similar art in mijorney. The example below is the art I created in Photoshop without using neural networks:
