USING AI FOR WRITING AND ART

USING AI FOR WRITING AND ART
When most people think about Artificial Intelligence(AI) and Machine Learning(ML), I imagine science-fiction and robotics come to mind. Sophia, a robot made by Hanson Robotics, went viral when she was first unveiled in 2016. She appeared in interviews on television shows like Good Morning Britain, 60 Minutes, and The Tonight Show, and she became the world’s first robot citizen when she got Saudi Arabian citizenship, raising questions about whether she would be allowed to marry or vote. [1]
Boston Dynamics went viral when they started showing videos of their robots that can pick up objects, navigate terrain, avoid obstacles and even stay balanced when pushed and kicked[2]. Their humanoid model, Atlas, can dance and do backflips[3], often leaving viewers either impressed or uncomfortable as we are reminded of all of the sci-fi horror movies or TV shows about AI we grew up with. Their dog-like model, “Spot”, even became one of the inspirations for the Black Mirror episode “Metalhead.” This episode, though frightening, feels more like a plausible “worst-case scenario” rather than an imaginative, far-away sci-fi future. Robot drones, for example, have already been used in warfare. 
But AI/ML is a lot more interesting than just creepy robots: AI is already a big part of our lives. Companies like Google, Youtube, or Twitter already widely use AI to moderate content on their platforms. Facial recognition algorithms are now used to unlock phones. AI is quickly getting more powerful and essential. And to me, an unexpected development happened in AI: they became creative. Algorithms like StyleGAN and NST (more on those later) show how AI can actually be used to create art. Google’s AI experiments have some interesting artistic applications as well, such as creating new instrument sounds by combining two other instruments with NSynth. For instance you can combine the sounds of a harp with a violin to get a strange, unique in-between sound. You can even use animal sounds, for example mixing a dog barking with a trumpet. Maybe you want something more playful, like an AI-karaoke that rates how closely your singing can match the voice of Freddy Mercury. [4]
But, perhaps the more interesting AI applications are language models: Language models like OpenAI’s GPT-3 can have incredibly human-like conversations[5].  Strangely, GPT-3 can actually excel at tasks it was never intended to perform. It can even generate fully working code in languages such as JavaScript, Bash, and React[6]. It is only a model trained for language, yet it seems to have learned other skills such as programming, basic math, and arithmetic. It seems that GPT-3 is so intelligent, it has actually learned how to learn. GPT-3 is so good at writing cohesive text that the editors of the Guardian had GPT-3 write one of their articles. According to these editors, GPT-3’s work was even better than its human counterparts:
“Editing GPT-3’s op-ed was no different to editing a human op-ed. We cut lines and paragraphs and rearranged the order of them in some places. Overall, it took less time to edit than many human op-eds.” [7]
Here are some examples of how I like to use AI (mostly for creativity) and how you can get started using AI right now:
Teachable Machine
Creating your own Machine Learning model is becoming increasingly less difficult: Google’s “Teachable Machine” is an easy way for non-programmers to create classification models based on images, sound, or pose. All you have to do is upload some training data, label them, and process them. I managed to create a model that can detect whether my hand is closed or open in less than 5 minutes. Programmers can even export the created model and use them in their own AI projects. [8]
StyleGAN and NST
I always imagined that creative jobs like painting, animation, or writing music could never be enhanced or replaced by AI. I now believe that I was wrong. Algorithms like Neural Style Transfer(NST) or StyleGAN show how easily AI can already generate completely unique artwork or even completely copy the style of an image and apply it to another. Do you want to see what Da Vinci’s “Mona Lisa” would look like if it was painted in the style of Van Gogh’s “Starry Night”? DeepDreamGenerator.com lets you create art with existing images using NST.
Blog asset
Do you want something completely new? Use Artbreeder.com to generate completely new faces, landscapes, or objects and use sliders or “cross-breeding” to create variations using an online StyleGAN implementation. If you need a bit more control over the layout of your landscape: Nvidia’s GauGAN can turn images that look like something a child would draw in MS-Paint into realistic-looking landscapes.
Blog asset
Sound maker (NSynth)
NSynth(Neural Synthesizer) is an algorithm for generating sounds.
Sound Maker is a program that uses NSynth to create new sounds by interpolating between two sounds. For example, you can create an instrument that sounds like it’s somewhere between a harp and a guitar, rather than sounding like a harp and a guitar-playing at the same time. They even included some animal sounds in their set of instruments, so you can hear what an instrument would sound like if it was a mix between a guitar and a cat or a trumpet and a dog.
OpenAI’s language networks (GPT-3)
OpenAI is a company that’s working on several ML models. One of their most well-known and most impressive models is their language model, Generative Pre-trained Transformer 3 (GPT-3). GPT-3 is impressive because it is a model only trained to learn language, and it became so good at understanding language it started to actually show signs of complex understanding and intelligence. It taught itself how to code, do basic math, and even generate simple drawings. Lazy programmers will no longer need to document their code: you could feed your code into GPT-3 and let it generate the documentation for you. GPT-3 is currently still in a closed beta, meaning you’ll need to send an application before you are able to access GPT-3. However, there are already some very interesting applications that are publicly available. Some of my favorites are mentioned below:
AI Dungeon 
AI Dungeon is a text-based adventure game. Text-based adventure games are one of the oldest types of games. However, these games are usually pre-programmed to send certain responses that are already written by the developers, only when a user uses commands with certain keywords that are again pre-programmed to be recognized. A Dungeon, initially built on GPT-3’s predecessor GPT-2, is free from these restrictions. Users can be completely creative in their actions and can explore a world that is much bigger and more detailed than any team of writers and developers could create. Rather than exploring pre-written stories that are the same for every user, you can explore a completely unique and personal story every single time. This has enormous implications for future role-playing games: imagine a game like World of Warcraft or Guild Wars, where every adventure and interaction is unique to the player’s experience, instead of every quest and story being pre-defined and identical for every single player.
AI Channels
“AI Channels is a social network platform that enables people to collaborate with artificial intelligence agents along with their friends and co-workers. You can have a conversation with an AI that’s an expert at recommending movies or participate in a round table discussion with the greatest minds in history and ask a virtual Albert Einstein to explain relativity.”
AI Channels seems like one of the most interesting and, at the same time, one of the more surreal and perhaps creepy applications of AI I have ever seen. Imagine being able to ask a virtual Einstein to help you with your physics homework. Maybe you can ask virtual Madonna to judge your songwriting. These scenarios would previously be considered ridiculous sci-fi concepts, but AI Channels has seemingly already made this a reality.
Algolia
Algolia is a search engine that allows you to actually use language to search for things rather than keywords. For example, imagine reading the Wikipedia article on bread, and you want to know where the explanation is for why bread is fluffy: usually, it would be pretty difficult to find that exact paragraph unless they happen to use the exact words you’re looking for. Algolia allows you to find this paragraph based on its description, even if the word “fluffy” is never used.
Koko  
Koko is an online mental health intervention that teaches its users to help each other and think more positively. It can also detect if its users are in a crisis and will send these users through the proper channels to manage these crises: 
“Individuals at risk of self-harm and suicide are identified programmatically, via deep learning models. Compared to the status quo, our outreach methods result in a 23% increase in the rates at which users contact crisis services.”
Shortly
Shortly is a service that allows you to get a long-form text from short inputs. For example, you can give it the command “write me a short blog about AI,” and it will generate a completely new output. This can help writers overcome writer’s block or save time. 
To find out more examples of what AI can do today and what kinds of things it could do in the future, these resources can help:
    Two Minute Papers frequently showcase research papers about the progress we’re making in AI.
Papers With Code, is a collection of AI research papers that include links to GitHub repositories that includes relevant code so readers can immediately test the methods for themselves.
As you can tell, incredible progress is being made in AI. Other than being a great tool for data analysis, AI can be an incredible tool for creatives like musicians, digital artists, and writers as well. Though some developments in AI might be somewhat frightening, there are a lot more that have incredible implications for quality of life, productivity, and creativity.

Sources:
Author's experience:
    Timothy Wolfram, Unity Developer
[1] Wikipedia article on Sophia the robot: https://en.wikipedia.org/wiki/Sophia_(robot) 
[2] Boston Dynamics - “What’s new, Atlas?” https://www.youtube.com/watch?v=fRj34o4hN4I
[3] “Every time Boston Dynamics has abused a robot” https://www.youtube.com/watch?v=4PaTWufUqqU 
[5] “What It's Like To be a Computer: An Interview with GPT-3” https://www.youtube.com/watch?v=PqbB07n_uQ4
[6] “GPT-3 And Code Generation — AI-enabled Instant Software Development”: https://becominghuman.ai/gpt-3-and-code-generation-ai-enabled-instant-software-development-270795077cbd
[7] The Guardian’s article written by GPT-3: “A robot wrote this entire article. Are you scared yet, human?” https://www.theguardian.com/commentisfree/2020/sep/08/robot-wrote-this-article-gpt-3
AN INTERVIEW WITH OUR FOUNDER
Meet our founder, Michiel de Graaf. Find out why he decided to start Rebels and what motivates him to Realize Revolutions daily.
THE MAGIC OF 3D ART
With multiple programs out there, I decided to elaborate on my personal favorites, combined with those most popularly used
GAMING OF TODAY
What comes to mind when you think about gaming?
USING AI FOR WRITING AND ART
Did you ever imagine creative jobs such as painting or writing music could be enhanced or replaced by AI?