Making sense of Dalle-2 (The AI that can make paintings)

The relationship between AI and human beings is like that between the butcher and his sheep. The butcher comes to kill one sheep and that sheep makes noises, but the other sheeps watch unconcerned. Untill some other day, the butcher comes for the next sheep, whose turn it is now to make noises.

An AI made this painting, within seconds, from word prompt- 'farmer with axe and herd of sheep'

When the AI made calculations easy, the human calculators complained, but others did not care too much. Then the AI started playing Chess and Go very well. The chess and Go experts struggled to make sense of this, but others did not care. Now the AI has started making paintings. Many artists are struggling to process this development. The wider world at large doesn't care much.

The above painting was made by an AI called Midjourney. Within a minute. With just the prompt- 'farmer with axe and herd of sheep'. This and similar AI like Dall-e 2 can give you stunning artistic results with simple word prompts. Now everyone can make stunning artwork using these tools.

Given all this, what is the future of art? Where are we headed? Is this the beginning of the end for art?

The existential crisis kicks back in. Again, painting made by AI. Prompt- 'Existential crisis'

What is DALL-E 2? Is it an artist? What is art?

I tried to define art in this previous post as- 'Art is a means to store and express thoughts'. If we go by this definition, then can we call DALL-E 2 an artist? The answer to this question would depend on whether Dalle-2 can have thoughts. If we assume that for now that DALL-E 2 cannot have thoughts, then it can be instead seen as a tool that can be used by people to express themselves.

Just like photography. But unlike photography, DALL-E can give you visual outputs that are not actually out there. It ‘understands’ what a chair is, what an angry face looks like, how people walk, what clouds are, etc. With the help of word prompts, it can stitch all these elements together in a proper manner, giving the imagination of a user an immediate access to infinite visual possibilities.

Just a note. DALL-E as of now is not perfect. It has problems giving you output when your idea is complicated enough. But looking at what it has achieved, I have full faith that AI can become much better at this task. The worst case scenario for humanity would be an AI that will become akin to an artist with godly powers such that you go to this artist, explain it an idea for a painting, and the artist asks you questions and doubts where it did not understand what you are trying to explain to it, and you have a back and forth conversion at the end of which the AI fully understands exactly what you want in the painting and gives you the output within a minute.

Such an AI will essentially be self aware. Will it even need prompts at that stage? This line of thought opens a different can of worms altogether. We will not get into this now.

So, an artist using DALL-E to express themself would be like a director instructing the actors in order to get the desired outcome. A word prompt based artist would hence be practising a different art form altogether. 

Hence, just like photography, the advent of DALL-E should be seen as a new form of art. A new form of art that fights for space with painting! Because where photography was strictly restricted to quickly capturing the beauty of the outside world, DALL-E can also visually capture thoughts!

Is this the end of paintings?

What we are actually asking is, as a form of art, can paintings take you somewhere in the landscape of expressions where it is difficult to go using DALL-E as a tool?

The only hope for paintings that I can think of emerges from the very core of why painting exists as an independent artform in the first place. It can take you places where words cannot easily go.

As I have discussed in this earlier post, it is not exactly true that one form of art is inherently distinct from another. It all comes down to efficiency. You can always think of any painting as a sequence of 0’s and 1’s and then call that sequence a ‘word’ that successfully describes the painting. But human beings cannot appreciate the beauty in that sequence of 0’s and 1’s, and this is not necessarily a bug.

When we are hence talking about efficiency, the question reduces to- are there paintings for which it would be more efficient for me to just take the time to paint them rather than attempt to describe it to someone who would then paint it in a minute? Do such paintings exist which are easier to paint and see than to explain?

I think this is the only way that painting as we know it, is going to survive the AI revolution. Artists need to paint things that are too complicated or abstract to be translated into words. 

In principle, this is possible. I mean, it does not make sense to me to talk about an AI that gives you music as output from word prompts. The landscape of expressions for music is so well departed from the landscape of expressions for words.

I think the following artwork of mine can serve as an example of paintings that are difficult to translate in terms of words. The process of first explaining these ideas properly to another artist (or future DALL-E) and then having them painted would be so tedious for me that I would rather paint them myself.

Caption this

Words are there for things that are already there. But paintings can bring things into being that are not there. Sure, we can later develop words to describe the things that the paintings brought into being, but this points at the fundamental limitations of words.