DALL-E: Prodigy or Plagiarist?
(Time to read-4 mins)
Why blog about AI (Artificial Intelligence) tools?
This blog is about DALL-E, another
AI-driven tool. It enables you to create digital art.
This is my third blog in a row about
AI. In my previous blog ‘Lipika, O My Darling!,’ I wrote about ChatGPT, the AI-driven
text-generator; and in ‘Talking with the Dead’ about Character.AI, a tool which
enables the user to chat with various characters including the dead and gone.
Why am I blogging about AI tools? Do
the AI companies pay me a commission to promote their merchandise? No, they
haven’t yet heard of this author, and have top marketing professionals in their
payroll. Why then do I blog about AI?
Because I learnt about it recently,
and wish to share it with my readers. After reading my blog on Character.AI, Ashok,
my friend tried the tool and messaged me: Just had an amazing chat with
Socrates!
Painting by Prasanna x DALL-E
I had read about DALL-E, the OpenAI tool that can create amazing
digital paintings from natural language inputs. Recently, I tried it.
I asked for “An oil painting of a Royal Bengal Tiger in M.F.
Hussain style.” In a few seconds, DALL-E created the painting I had ordered.
Here it is:
Lead Painter
DALL-E is rather generous. It has
acknowledged me as a co-creator or co-painter, since I provided the idea (the
text for the image idea). Magnanimously, it mentions me as the Lead Painter. That’s
a huge massage for my ego! I will never be a painter on my own, but being a 50%
collaborator with AI feels pretty good. However, in my humble opinion, the
Credits should be: M.F. Hussain x DALL-E x Prasanna.
What’s DALL-E?
A free-to-use (with some limitations) digital
art creation tool, launched by OpenAI, a company which is focussed on
Artificial Intelligence (AI) research and deployment, and has a partnership
with Microsoft. It generates images
from text descriptions, using a combination of neural language processing and
image generation techniques.
DALL-E is a mash up of the words
Salvador Dali, and WALL-E. Salvador
Dali (1904-1989) was a renowned Spanish surrealist artist. WALL-E was a 2008
computer-animated science fiction film produced by Pixar Animation Studios and
released by Walt Disney Pictures, and is among the best animation films ever
made.
On 28 September 2022, DALL-E 2 was
opened for all. In Nov 2022, OpenAI mentioned that more than 1.5 million users
are generating more than 2 million images per day.
What can it do?
It can generate a wide range of images, from
photorealistic to highly stylized, based on the text input provided. DALL-E can generate images of objects,
animals, scenes, and more.
Based on Natural Language inputs, it
can create ‘imagined’ and fantastical digital art. You may ask: Create an oil
painting of a gorilla enjoying an ice-cream cone on Eiffel Towers, and DALL-E
would oblige.
What it can’t
It is not able to understand or
generate images based on real-world context or physical constraints, and the
quality of the generated images can vary widely.
Why did I Sign Up?
At school, I was drawing-challenged,
and was quite jealous of Guru, Shankar, Mahadev and other classmates who excelled
at drawing. For my scholarship exam at Class VII, I was required to pass a
drawing paper to qualify for the merit scholarship. I had scored 17/50! I guess
the examiner had been very generous to award me pass marks since the brinjal I
had drawn looked rather like a potato even to me.
But I had many superb ideas for
drawing and painting, and hoped to someday ‘learn’ how to paint. When I met
DALL-E, I sensed an opportunity to indulge in my fantasy.
What did I ask it to make?
I admire great painters, since
knowledgeable folks admire them. Being a novice, and having little knowledge of
painting, who am I to question the considered opinion of art critics?
I have gawked with awe at the amazing
paintings: ‘The Sistine Chapel Ceiling’ at the Vatican, and ‘Monalisa’ at The
Louvre. I also know the names of the famous painters of the world, and of
India. I could reel off their names, but this is not the place to show off my
general knowledge.
I asked DALL-E for the following:
· A sunset in Pachmarhi, MP in Van Gogh style
(You may read, if you like, my blog: Pachmarhi: Two Paintings)
· A sunset in Pachmarhi, MP in Claude Monet
style
· Sunrise at Bhopal Lake in watercolour
· A high-resolution image of a swarm of dolphins
at Bhopal lake (There are no dolphins in Bhopal lake!)
· Khajuraho temples in style of Rabindra Nath
Tagore
· A Royal Bengal Tiger in M F Hussain style
· Men hunting a bison in Bhimbetka rock cave
painting style
· Bhopal skyline in Gond painting style
· DSLR quality photo of a royal Bengal tiger
couple at the Taj Mahal
· Monalisa in Raja Ravi Varma style painting
· DSLR quality photo of Chilika lake on a bright
sunny winter morning with fishing boats, migratory birds, a group of cavorting dolphins,
and foreign tourists on a cruise stunned by Loch Ness monster towering over the
lake at a distance
· You may note that I began with modest ‘prompts and DALL-E produced in a jiffy excellent output, as ordered by me. I got bolder, and raised the difficulty level by adding multiple and complex parameters. DALL-E didn’t even raise an eye-brow, and went about valiantly and resolutely to deliver. AI is not allowed to be emotional, I think. My last ‘ask’ was indeed tough, and the result was average.
But why blame DALL-E? I had included several parameters about which DALL-E has not been trained yet. If I provide the feedback of what lacked in the output, it will understand my concerns and suggestions, and next time it will do better. It is capable of deep learning. Power of AI is not to be underestimated.
My Assessment
Quite impressive. My Rating: 7 on a
scale of 1 -10. But don’t take my word for it. Go ahead, Sign Up, try it, and
find it out for yourself.
Prodigy Painter or Plagiarist?
A painter will possibly dismiss DALL-E
as a plagiarist who shamelessly steals from the work of serious artists. Much
as writers may consider it beneath their contempt to even explore the
possibilities of ChatGPT.
A painter takes a day or two,
sometimes a week or longer to create an art. Da Vinci took about three years to
complete Mona Lisa. DALL-E works fast and furious and completes any assignment
the user gives in a matter of seconds. If you don’t like the output, it can
offer six variations at once. If you wish to tweak your parameters and ‘prompt’,
it doesn’t complain. In fact, it is raring to go.
Is DALL-E a prodigy painter, or a shameless
plagiarist? Since the tool has been released for all only recently, it is too early
to pass a verdict.
How To?
Go to OpenAI, and sign up for DALL-E.
Try it. It is not an App, but available at OpenAI’s portal. Here is the link: https://openai.com/dall-e-2
Yes, DALL-E 2 is the newer, improved
version.
Of What Use is DALL-E?
For me, to marvel at the capability of
AI, and the inventiveness of the engineers and software makers who created it.
For professional artists, it offers immense opportunities to be super-creators,
by intelligent use of the capabilities of DALL-E. Possibly, it would be of commercial
use, too. Lastly, a painter could co-create with DALL-E, say on 80:20 (Human:
AI) basis or any combo that she prefers. For low-paying assignments, she could
even quickly make a 20:80 art, and the buyer may hardly note the difference.
***
- Author's profile may be seen at http://amazon.com/author/pkdash
- Books by this author are available on Amazon.in, Kindle eBook, Flipkart, and Notion Press, Chennai.
***
Nice to read this article...educative one about AI...
ReplyDelete