Eye Suriyanon

Thai-born, Glasgow-based artist and sound designer.

Conjuring future voices for sentient mechnical beings based on Buddhist Philosophy, Spectre and the in-between through Moving Image, Machine Learning & Spatial Audio.


Understanding Artificial Intelligence & Machine Learning

AI is everywhere.
What is it? Will it take over? Can it think?

This post is your quick guide to grasping the main concept of AI:- what it is, how it is made, what it is for and how we can use it responsibly.

︎ What is it?
AI, is short for, Artificial Intelligence, a term used to describe a computer system or programme that is designed to appear smart like a chatbot. The programme is constructed using machine learning, a process which allows the system to learn and adapt with little to no instructions based on finding patterns from a large collection of datasets. The information gathered in these datasets is often generated and moderated by us - humans.

︎ Will it take over?
AI is a programme that lives and operates in our computers and servers - just like any other programmes we use. Thus, it is a tool.

A tool that has its limitation and is built from our imagination, therefore made from the human mind so its capacity is limited by human intelligence. If we can’t make a sentient machine right now, it means that AI is not sentient. It cannot function without us giving it instructions.

As someone working with this tool, it is important to remember how the tool is being represented. The glorification of its ability perpetuates the myth of ‘machines overtaking’ or ‘sentient machines’ ever further. 

︎ Can it think?
The function and capacity of AI are often misinterpreted as machines being able to think freely. This concept is actually called ‘Artificial Consciousness’ or AC, which refers to a human-created machine that is aware of its own existence and can think like it has a mind. AI and AC are not mutually exclusive. AI is a tool and AC is a theory.

︎ Feel free to do your own research ︎

AI Now Institute: a research institution highlighting  the ethical and environmental issues surrounding AI.

In Machine We Trusta podcast by MIT Technology in Review  explores the impact of AI on our daily lives.

Interdependence: a podcast by Holly Herndon and Matt Dryhurst exploring the use of AI and other new tech by creatives.

ML4A a website dedicated to the creative use of machine learning with free educational resources  by Gene Kogan.

︎ How to use it responsibly?

As the tool functions in the digital realm, it can often feel disconnected or unreal due to the lack of physical impact the creation of the work has.

In reality, working digitally is not as sustainable as you think. There’s no right or wrong answer to this question. All I’m asking is for you to be considerate of the ethical and environmental impact digital production has.

  1. Hardware - what is your computer made out of? The compounds are extracted from the Earth through extensive human and machine labour, refined in factories fuelled by the planet, and combined with chemicals, transported by carbon-producing vehicles. Where possible, please opt for secondhand or refurbished hardware. This helps to reduce the amount of electronic waste as some components are not recyclable.

  2. Power - How are you powering this hardware? And how energy efficient is your shiny 4K widescreen?

  3. Software/ Programme - Who are the developers and who funds their research? And where are they getting the data from?

Read more about it 
Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence by Kate Crawford
The book reveals how the global networks underpinning AI technology are damaging the environment, entrenching inequality, and fueling a shift toward undemocratic governance. She takes us on a journey through the mining sites, factories, and vast data collections needed to make AI "work" — powerfully revealing where they are failing us and what should be done. Atlas of AI was named one of the best books on technology in 2021 by the Financial Times.

I have too many to remember to add in this section. Plus, this is not academic text please don’t quote me.

Now that we’ve addressed the big three questions, let’s move on to how Machine Learning works!

Machine Learning is essentially what gives AI its intelligence. It is a process that teaches machines how to learn. Like, how we learn through associations, recognising patterns or through trial and error. These methods of learning are called classifications in computer terms, by using an example dataset to build its own definition of things so that when new data gets shown it can work out what the data is and what it can be used for without a person having to enter the programme to define the data. 

  • Supervised Learning is like learning by association. It learns by matching up either a word with an image or vice versa. The computer sees these as two types of data that belong in separate groups, to match them up, they need to share a common thing. i.e . a word needs to have an image and an image needs to have a word attached to them. Therefore all the data needs to be labelled and have images. The first few examples of the matches are done by a person and then it is a big game of matching by the programme.

  • Unsupervised Learning is like recognising or finding patterns by looking at them without a person telling them what to look for. In the dataset, there are words with no image, images without words and images with words. Then it starts to create groups that have similarities to each other:- shares the same word or the colours or the outline etc. Then repeats this until everything is grouped.  

  • Semi-Supervised Learning is a mixture of Supervised and Unsupervised Learning, this helps to increase learning accuracy when dealing with large volumes of unlabelled data. This means it starts with us matching a few things and then finding matchings by itself through recognising patterns.

  • Reinforcement Learning is like taking a quiz that gets marked. But you know haven’t studied for the quiz so you guessing what the answers are. Luckily, your examiner is really nice and allows unlimited re-takes until you get all the answers right. A mixture of trial and error, with a person that rewards you when you get an answer right but punishes you when you get it wrong. The examiner is a person so the correct answer may differ. 

If you’d like to learn more about Machine Learning 
I’d highly recommend the Machine Learning with Python course by freecodecamp. With short videos with simplified delivery, the course is great for those with a foundational knowledge of Python. - this is free!

Fancy building your own model? 
Google offers a Machine Learning Crash Course with TensorFlow API. It’s more in-depth and requires coding know-how - absolutely free!

︎ What does this all mean when using it in a project?
This means the time it takes to complete a process can be shortened when prototyping an idea or experimenting with outputs.

︎ What is it being used for?
The implementation of AI for the creative field tends towards being generative, i.e. being used to create something new based on a prompt. This type of machine learning is called GAN, which stands for Generative Adversarial Network. In the most crude explanation, it learns to generate new data based on the given statistical model. Similar to mimicry in evolution biology where a specie evolved to resemble another organism or object to protect itself from predators. So, the new data isn’t a copy-paste job, it’s a copy-mix-paste job.

︎ Here are some existing ML-GAN models ︎

︎ Text-to-Image: VQGAN+ (free), MidJourney, DALL-E
︎ Video Generation: RunwayML
︎ Audio Generation: SampleRNN
︎ Music Generation:

WARNING: Use them with caution. Be aware of where the datasets are from - are there unauthorised work by others?

Created by Eye Suriyanon, 2023 on Cargo Collective