Ayo AI

With great power of information technology comes great responsibility.

The pace of progress of generative AI is faster than any new technology in history. Even AI gurus and scientists are worried, and governments are scratching their heads about regulation.

The image of Kathmandu Valley is an example of AI art, and we have other illustrations of what this technology can do, including producing a prompted promotional video for Nepali Times in both Nepali and English. 

Nepal is gearing up to join the AI bandwagon with user-friendly apps becoming more advanced than Tiktok, Facebook, Instagram. Many young Nepalis are learning and adapting quickly. Parents and teachers now need to pay attention to guide children to use the apps safely.

PROMPT: Glass morphism, high quality, ultra-realistic huge glowing cactus, night in a desert, ultrahigh quality, extremely detailed, matte texture, pop surrealism.
PROMPT: Nepali mother holding a baby, with volumetric dramatic studio light, sparkling garden, silhouette, extremely detailed, super resolution. Desert punk, sun light at night, with glittering tree

 The number of apps grown from just 3 generative AI apps in 2015 to 33 in 2022, and in just five months of 2023 until May, 640 new apps have been created.

COVID-19 pandemic was a key factor in accelerating AI growth, as businesses adapted to remote work. We are seeing the rise of AI apps that can transform text to high quality, ultra-realistic photographs, art, videos, audio, voices and much more are coming. Most are only in Beta versions and their full potential is yet to be unleashed. That will not take long.

The apps also have the potential to help not just the entertainment and advertisement agencies but also health and education sectors in Nepal. 

Computer machine and info tech experts around the world are worried about the potential risks. AI-generated contents can be used for ‘deepfake’ videos and images, fake identities and spear phishing. 

PROMPT: Graphic art, ultra-realistic, robot is painting, 3D 8k HIGH quality illustration of a mourning young woman watching AI powered bot drawing on a canvas under Fiber optic lighting, Halogen lighting, Jade color, insanely high resolution
PROMPT: Painting of many Viking warriors fighting in the battlefield in insanely high definition,  with chartreuse color with photography lights and dark lights opposite a blue ocean and Viking ship

The apps can also clone faces and mimic a real person’s voice that can be used for spreading misinformation and political propaganda, or be used for blackmail and manipulating  public opinion at election time.

These incidents are quite rare now but experts are asking governments to take steps in developing and implementing ethical guidelines for using these AIapps. European Union has released guidelines for the use of AI technologies in March 2021.

Geoffrey Hinton, known as the ‘godfather of AI’ is one of the most powerful voices sharing concerns about the misuse of deep learning to spread disinformation.

AI and I

Let's suppose Leonardo da Vinci was alive, and he agreed to paint a masterpiece for me. As a Samurai fan, I would tell him to paint a super detailed, ultra-realistic artwork of two female Samurais with sullen expressions under volumetric light with matte texture. In just five seconds, da Vinci creates a spectacular piece beyond my wildest imagination (Image 1).

The only thing is: it was not da Vinci who painted that image, but http://leonardo.ai. That striking portrait of two Samurais was created on my laptop by the advanced generative Artificial Intelligence (AI) application that creates images from natural language descriptions, known as ‘prompts’.

PROMPT: Create a full body graphic art of 2 female samurai with sullen expression, silhouetted and volumetric light.
PROMPT: Create a hyper-realistic, 8K close up photograph with a cinematic view of a young Nepalese mother holding her baby.

Leonardo AI converts text into any kind of image: painting, pencil drawing, cartoon, photograph or diverse art forms. It is one of many advanced AI-powered image generators that have come up in recent months. Since its release in February, Leonardo’s users have increased from 1.5 million to 11 million and on 1 May, it was released to the public. I got early access and since then I have created nearly 5,000 images.

There are other advanced image generation applications like Midjourney, DALL-E 2, Stable Diffusion, Lexica Art and BlueWillow. One of the latest entries is Microsoft’s Bing Image Creator, released publicly in April. 

Some of these programs even provide subscribers with a copyright licence to publicly share and sell art online as NFTs (non-fungible tokens sold as unique digital assets and proof of ownership). Some of the biggest online stores like Etsy, Open Sea, Rarible, Mintable etc have offered services to sell AI art as NFTs. 

Text-to-image generation is not new. But recently there has been an explosion of user-friendly, low-cost and free generative AI apps that have taken the world by storm. 

PROMPT: Hyper realistic super resolution, full body, wide lens photograph of young high cheek bone Nepali, brown skin tone, holding a white glistening Yak, wearing mountaineering outfit with Chiaroscuro dramatic studio light, night, silhouette, extremely detailed. Pretty eyes, sparkling Mount Everest and glistening moon.
PROMPT: Award winning photo of a Nepali girl, 35mm, F/2.8, insanely detailed and intricate, character, hyper-maximalist, elegant, ornate, beautiful, exotic, revealing, appealing, amative, hyper-realistic, super-detailed.

A game changer

On 11 May, Google’s new chatbot Bard was released in 180 countries, and dropped the ‘waitlist,’ which means that people do not have to register anymore for access. One of Bard’s notable features is Adobe Firefly, which is integrated into its web app and is already predicted to be better than other apps. What makes it different is that the app models are trained on openly licensed stock images and user-generated contents.

Although this will be a game changer, artists have started to question the ethics of text-to-art generators. Like chatbots, arts generators are also AI-powered and trained on massive datasets of images and texts. 

Low-intensity warfare has already broken out as artists identify their niches, styles and trademarked characters in AI-generated images. This is not yet an issue for Nepali artists who are not concerned that AI art will replace them. 

“Artists have survived technological advancements. The medium has changed but the passion, imagination and creativity process of artists have always remained unchanged,” said Kailash Shrestha, contemporary visual artist and initiator of Artudio.

Some artists in the USA have even started suing AI companies for copyright infringement. In January, a group of women artists in California filed a case against Stability AI, Midjourney and Deviant Arts for stealing intellectual property. In February, Getty sued Stability AI for infringement for copying more than 12 million copyrighted images without permission.

PROMPT: Ultra-realistic, insanely high-resolution satirical cartoon of global warming for a newspaper.
PROMPT: Make a hyper-realistic Gurkha warrior with Chiaroscuro dramatic studio light, night, silhouette, extremely detailed, super resolution. Sparkling Mount Everest and shining moon.

The companies asked the courts to withdraw the cases arguing that AI-created images are not similar to their work, and did not note any specific misused images. There has also been a backlash from generative AI users’ community against critics. 

A Twitter war is going on between artists and AI app users. But only a handful of artists are raising their voices. There are now already 300 generative AI programs and apps, and many of them are free to use. Despite appeals from some artists to be sensitive and join them in their fight against intellectual property theft, there has been little response. 

The generative AI arts community is expanding by millions every month with the majority of users from USA, China, India and Brazil. In Nepal most AI users are using Midjourney.

Calling themselves AI ‘artists’, they have created their own networks, online forums and on social media platforms especially Instagram, some have become AI stars. YouTube has become the preferred platform to share AI art, creative ideas and tips for improving their text prompt skills. There are already books on text prompts which are flying off the shelves.

Many users are selling their work, even though most are doing it more for fun than commercialising the creations, according to a 2023 poll by Midjourney which showed that 70% of members never shared images for commercial purpose, and a majority do not have backgrounds in art or design. Over 83% use it as a form of art therapy, and mostly posted on Instagram to share within their own community.

PROMPT: Graphic art, portrait of old gorilla, Fiber optic lighting, Halogen lighting, Jade color
PROMPT: Graphic art, portrait ultra-realistic, mourning mother carrying daughter, facing front, Fiber optic lighting, Halogen lighting, Jade color

How do I join?

Ethics: Despite criticism by artists they do not oppose this innovative technology, but they are frustrated with Big Tech stealing intellectual property. While using any app, read the terms and conditions before publicly sharing or commercialising anything.

AI Tools: There are plenty of videos on YouTube that provide tutorials on AI art generators. Microsoft’s Bing Image Creator is one of the most user-friendly. Mage Space, Lexica Art are also simple to use. Leonardo AI gives a lot of tokens without a paid subscription, and I created all the images on this page using it. These images are just for educational purposes. There are tons of other apps on Future Tools.

Be Creative: Prompt texts creatively. Don’t just write ‘monkey’ or ‘dog’. Get to know artistic concepts, especially styles and techniques. Study and learn the parameters: lighting, types of cameras, colour palettes, dimensionality. Describe objects, people or scenes. Make your own list of different photo lightings, colours, style, textures, moods, expressions etc. Prompt Hero gives ideas and examples.

Join a Community Platform: Create a Discord account.  To use other tools like Midjourney and BlueWillow, you have to register with Discord anyway. Niche List: Create your own niche to make images stand out. For ideas, join chatbots like ChatGPT, Bard, Bing Chat and Hugging Face.

AI generated promo video for Nepali Times

Chatbot generated script to voice narration and model to create an AI generated promotional video for Nepali Times. Edited via CapCut. Generative AI videos by NARESH NEWAR

 Create a video with a generative AI friend to do the same video in Nepali through the beta version of https://www.kreadoai.com. Sign up with Google and click 'oral broadcast video creation' on the dashboard. Chose language (Nepali is also included!) language style (there is also a shouting version), character and model. Translate text with chat.openai.com. Bard doesn't speak Nepali yet, but it apologised and promised it will be included soon, haha. The app won’t let you generate unless you have listened to the whole thing. Generation takes some time and AI will email you once it is done.