Blog + Essays
I'm learning Mandarin: Why I think AI agents are the perfect language learning partner
Since learning Chinese I lack opportunities to practise speaking the language. Speaking practise opportunities are one of the biggest roadblocks to learning a new language according to a 2024 study.
My vision: Walk around your own home practising speaking out loud to your agent, and have it roleplay speaking scenarios in the language you want to learn - like it was a real interaction. This is the closest thing you will get to a real native speaker and we now have advanced enough Agents to do this. This is a game changer. It not only democrotises language learning by presenting a cheaper alternative that tutors and classes but lowers the accessibility for anyone who wants to learn lagnuages in their own time. Personally, I've been doing this for 3 months now and my mandarin has definitely improved.
Agents make the perfect language speaking partner. It’s a perfect use case for AI.
Here's why:
Affordable
Compared to expensive tutors.
Patient
No need for feeling embarrassed or rushed.
Always Available.
Study in your own time in your own home.
Customizable.
Lessons are completely tailored.
It’s the best alternative for practising speaking with a native speaker.
Mock up Scenario - Walking and Talking in your kitchen
Learning on the go - multimodal possibilities
All That Glitters Is Not Gold: Is the language and graphics of 'Magic' used for AI tools...
1. Unhealthily disconnecting users from the fact that there’s a real, material process behind these digital functions — and that their facilitation has tangible environmental impacts due to the growing demand for data centres?
(It feels similar to the meat industry, where words like ‘pork’ disconnect us from the actual process.)
2. Perpetuating an unhealthy concept of limitlessness — because it’s framed as 'magic'.
3. Widening the gap between those who understand how it works and those who don’t, making people apathetically dismiss the idea of trying to understand this new AI phenomenon, as it seems like ‘magic’.
(Which is dangerous, given how important it is for society to grasp the implications of AI.)
Pancake Recipe: How LLMs are changing search and the landscape of the web
In this talk with Yosuke Ushigome, he makes a great point about how llms will change the very foundations of the internet and how we search for things.
Pancake recipe: It's a sunday morning and you decide to look for a pancake recipe. You would normally 'google' the recipe. But not you get a quicker and simpler recipe from chatgpt. Poor Ol' Recipe site gets no traffic.
I find myself now using chatgpt more often for mundane things as I'm frustrated with googling something like a recipe and the first things that come up are two 'sponsored ad' pages a half covered google ai recipe. and then finally a food recipe website on pancakes which I usually have to scroll past a whole story or pancakes to find the recipe.
What are the implications of this ? and will it entirely change the face of what the internet looks like and unsure wether for the best.
AI Democratises Language Learning by presenting an alternative to expensive tutors and classes.
Democratising Technology: How AI is democratising Language Learning
Learning a new language meant paying for private tutors, pricey courses, or premium apps.
But AI language models, like ChatGPT, are starting to change that.
Now, anyone with internet access can practise speaking, learn grammar, and get instant feedback — often for free. These tools can offer personalised, on-demand conversations, making it easier for people to learn at their own pace. It’s especially valuable for people in remote areas or those who couldn’t afford lessons before.
AI isn’t just convenient — it’s democratising language learning.
AI Forecast: Will real artwork become a sign of luxury for brands and companies ?
I was looking through wired amgazine and noticed a large amount of the illustrations were all generated. This caught me and my friend by surprise and couldnt get away from thinking how this came off as a cheap move by such a large company.
This makes me call back to the luddites, craft movement and industrialists in the early 19th century. Will the artisan or master of craft become a valued and purposefuly chosen commodity for companies in the future.
I was speaking to a lead designer at a tech company and he said that he was leaving the graphic design game because now work can be generated and iterated 1000x fold. Which just increases the amount to which people are worked. Not the quality of the work.
AI and Education: "How to know someone has learned"
*I have written this without the use of Open AI's ChatGPT
I went back to my school recently to give a talk on my journey as a designer and some of my work. I started talking to the head of science, we quickly got into a discussion about the emergence of AI and its probelms in Education. I said it transformed my final year of university, while he highlighted the minefield is has been for educators to navigate the students being far more able in this tech that their own teachers.
Inspired by VOX's journalism video on Education and AI and my talk with the head of science at my old school, I started a research project. I wanted to investigate some of the questions I found interesting - stemming from the statement if "ChatGBT can do your homework for you… " then
"how do we know when someone has learnt something" and "do educators have to cite the source of the AI now?".
AI 'tools' are integrated into nearly every professional software from Google Docs, Grammerly, Notion - it seems redundant to ban these tools in education if it will only disadvantage the student from adapting to using them later on in the work environment. It seems educators are at a crossroads of whether to integrate LLMs into the educational system or to ban them outright. I ask what measuring sticks must we now use to 1. know someone has 'learned' and 2. for educators to know they have 'taught'.
This list of goals I believe we should aim for.
1. Do not leave the responsibility to children to self regulate the tempting use of AI to pass the current education system.
2. To not add 'policing the use of AI' to the workload of teachers.
3. To not disadvantage students futures by not embracing the obvious rise and ubiquitous integration of AI in professional working world.
4. Teach children how to use AI to enhance their learning and not to indirectly teach them how to specifically enhance their ability to cheat the current education system using AI.
5. Prepare a new generation for the complexities of misinformation (or inaccurate cited information) created by AI and LLMS - by giving them them thorough critical thinking and digital observational skills.
6. Reduce the impact and effectiveness of misinformation per person by updating the current education system to
Giving a talk at my secondary school - 2025
Vox - AI can do your homework. Now what?
The 1980s China keyboard crisis
Historical Synopsis: After Mao Zedong died in the early 1980s new premiere Deng Xiaoping realised if China was to prosper it needed to join the growing computer industry. The only problem was how to fit a 50,000 + character based language on the western 80 keys QWERTY keyboard you are most likely using right now. The answer lay in an alphabet created 30 years earlier called Pinyin.
Flashback to the 1950's Mao Zedong wanted the Chinese language to be more accessible and raise the literacy of the country. He initially wanted to scrap the Chinese written language altogether for a Romanised system and was interestingly persuaded against doing this by Communist Ally Joseph Stalin to keep Chinese 'Hanzi' or Written Language.
However Mao Zedong still keen to promote literacy introduced 'PINYIN' in 1957 as the official Romanized system to translate Chinese (winning the small struggle of which of the multiple Romanised system would be used).
Pinyin is basically a phonetic way of sounding out Chinese words with English letters. for example;
我 is the character for 'I'.
我 when spoken, literally sounds like the sound "wo".
So you type on your english QWERTY keyboard the letter 'w' and 'o', to create the word 'wo'.
So 'wo' is how you say 'I' in Pinyin.
我 [Chinese character]
wo [pinyin]
I [English]
So in the 1980's the Chinese government pushed developers to create a system that uses pinyin on western QWERTY keyboards to construct Chinese characters. Thus saving the written language from digital extinction. The system they created was good but slow, typing in english letters to make chinese sounding characters, so in the 90s a predictive based algorithm was created to take the inital 'root' character of the pinyin word and predict what character you were typing
https://en.wikipedia.org/wiki/Cangjie_input_method#:~:text=Unlike%20the%20Pinyin%20input%20method,on%20a%20standard%20QWERTY%20keyboard.
Traditional chinese
using BPMF
Structurally different in construction.
new 20th century invented alphabets and languages of PINYIN and BPMF.
China 80S lagging behind west in computer - a national crisis of how do we break into the computer world with Chinese characters.
Anti-Homeless Architecture: Project Essay
https://drive.google.com/file/d/1y64qBAJlYhnX66yUn5L6dBiv1fIvlPZ6/view?usp=drive_link
A small written and visual piece on Anti-homeless Architecture and its effects to the public and those it is designed for.
"Hostile Design, is an intentional design strategy that uses elements of the built environment to guide or restrict behaviour in urban space. It often targets people who use or rely on public space more than others, like people who are homeless” - Cara Chellew, 2019, Canadian Institute of Urban Studies