As promise
English version of "Artitificial intelligence : friends or foes"
On YakiHonne (npub1yzv…rf8q)
Sorry for the delay upto 3 days after Thai version was out
In this article , I'm going to share with you my experience of having AIs as my personal assistant to do my work as a doctor and content creating including the conflict views of AIs and my thoughts on that
I heard about AI for the first time around the end of 2022 through online media. Back then, I thought it would be difficult for ordinary users to access it; only large organizations with influence could possibly gain access rights.
Until ChatGPT came along, offering free access before introducing paid packages. I immediately jumped in to try it out. I was amazed by its capabilities. If anyone remembers around the end of 2022 and the beginning of 2023, influencers, YouTubers, and content creators in the tech and self-development spheres were all talking about it—it became a viral trend.
To test ChatGPT’s abilities, I tried asking the following question:
Medical knowledge
I delved deep into medical journals and asked where the sentence came from and who wrote it. Some journals, especially older ones, couldn’t provide the answer directly but suggested similar references for me to explore further.
I inquired about the steps in performing CPR, a specialized skill of mine. ChatGPT provided a comprehensive response, succinctly emphasizing the crucial steps.
I sought opinions on management strategies, tweaking certain aspects in my hospital… (Currently, besides being a physician attending to patients, I also lead three major units.) The suggestions were intriguing; some I implemented into actual practice, resulting in tangible outcomes in the hospital. We managed to significantly reduce patient mortality rates of “pulmonary embolism” (or PE) , which was quite satisfying.
Once, a director asked for my input regarding terminology for patients who passed away before reaching the hospital. There were two terms: “out of hospital cardiac arrest” (OHCA) and “dead on arrival” (DOA). They wanted to streamline it to reduce data confusion for statistical analysis. At the time, I vaguely recalled that “DOA” had been phased out for a while, with “OHCA” being the preferred term globally within emergency medical systems. I excused myself to the restroom for a moment and used ChatGPT to clarify. It provided a clear explanation that while “DOA” might add confusion, “OHCA” was a standardized term used worldwide in emergency medical systems. I returned to suggest using only “OHCA,” leading to a summary and progress in our work. The hospital subsequently removed “DOA” from its documentation system.
Once, I posed an extreme question to test ChatGPT’s nerves. I let it present its capabilities directly to the hospital management, regarding how to effectively integrate it into our hospital for competitive advantage against private hospitals of similar caliber. Its responses were highly intriguing, such as setting up AI customer care to answer basic health inquiries without patients having to visit, automated text notifications for appointments replacing manual reminders, or even advanced programming to embed ChatGPT into our information system for statistical analysis, enhancing our main infrastructure.
Some of my peculiar questions those google doesn’t fit
Why do countries in the Asian region use bidets for cleaning after using the toilet, while Western countries use toilet paper, even reaching the point of decomposing toilets . Is it really better? Are there long-term consequences, especially regarding odor?
There’s interesting information suggesting that people are becoming more aware that using bidets at home is cleaner and causes fewer long-term issues. I’m waiting to see how this trend evolves.
Before the Federal Reserve of the United States gained such overwhelming power that it could control global financial affairs, which is the topic of my article on the Thai bitcoiners community named Rightshift, “What’s happening in the world… Why was Bitcoin created?” (in the future I will publish English version of it stay tune for update)
Helping with my article writing
The long-form content article on Minimalism that I wrote is a critique of the consumerist system that numbs people into excessive consumption, leaving them living solely to pay off debt. It’s the angriest piece I’ve ever written. What happened was, I wrote the final paragraph trying to find a breaking point to conclude the article but couldn’t. Then, I turned to chatGPT for help. I asked it to assist me in writing this article, with this kind of content, to help condense the final paragraph to wrap up the article nicely. I rearranged its suggestions to finish the work much faster.
The article “Minimalism,” I’ll publish an English version soon. It’s an article of mine that I believe is highly beneficial to those who read it and can’t be written elsewhere because the modern capitalist are not going to be happy with what I wrote.
Assisting doing business with Japanese anime fans
At the beginning of March 2024, I decided to start my small online business using generative AI to create anime-style cartoon images to market to Japanese anime fans. I’m currently learning Japanese, but I’m still at a basic level and can’t communicate fluently, so I rely on chatGPT to translate for me. Then, I study the phrases it translates to understand their sentence structure.
This is an example of AI using a large language model (LLM) for interactive conversation with users. Until now, I’ve completely shifted the way I work. Usually, when I search for information, I’d turn to Google. But now, I use chatGPT along with Perplexity AI as the main tools for information retrieval. It’s much faster and more precise.
Next, there’s another type of AI that I use.
Image generative AI
Or AI generates images from our prompts. After I experimented with chatGPT in early 2022, within a few weeks, I got to know Midjourney as the next step. It was the first image-generating AI I played with.When I tried it out, it was still free at the beginning. However, due to its rapid popularity worldwide, it probably caused the servers to work excessively, to the point where the developers had to restrict free trial usage and introduce monthly subscription packages instead. At that time, I wasn’t quite impressed, didn’t understand it much, but through perseverance, I got acquainted with Stable Diffusion.
Stable Diffusion (SD) served as my mentor, honing my skills until I became proficient in using AI image generation to this day.
The advantage of SD is that you can download and install it on your machine for free. It primarily utilizes your device’s resources for image generation processing. The quality and speed of image generation depend on the VRAM of your graphics card. With my 6-year-old laptop and GTX1060 4GB graphics card, each image generation takes 3-5 minutes, which is considered long compared to higher-spec machines. I joined the SD Facebook group and relied on asking questions, learning from others’ experiences, and watching tutorial videos shared by group members. But now, many friends have transformed and started offering AI image generation courses, which is another commendable professional growth path.
During that time, I spent over 10 hours a day practicing, more than playing games. I must say it was torturous for my laptop, and eventually, my graphics card started having issues. Whenever I opened multiple tabs or some websites with high-resolution images, the screen would flicker, blue screen and restart.
As time passed, newer versions of image generation models in SD became significantly higher in resolution. In just 2 months, my laptop couldn’t handle image generation anymore.
Hit the Generate button, blue screen , laptop restarts, and the loop continues.
During that time, there were developers wrote code on GitHub to install SD on Google Colab. After some trial and error, I finally managed to use it on Google Colab. The workflow became much easier. Previously, it took 3-5 minutes to generate one image. Now, with the initial resolution of 512 x 512, it takes about 10-15 images in the same time frame, allowing me to practice more and learn new techniques.
However, eventually, Google blocked the use of SD on Google Colab. To continue using it, one had to upgrade to Google Colab Pro. During that time, I began writing on platforms like Rightshift and Yakihonne. I distanced myself from image generation for a while.
Until one day, something on nostr introduced me to Pixai for anime-style image generation, which I enjoy. It has similar features and tools to SD. When I got my hands on it, I immediately knew how to use it. I also delved into the world of photo editing programs like Photopea and video editing with CapCut.
It turned out that from studying chatGPT, I got to know Midjourney, followed by Stable Diffusion, then Pixai, and branched out into other related programs like Photopea and CapCut.
Oh, ok bro! So, you’ve encountered LLM and image generative AI, huh? So … what makes them so special to you?
I see a wide range of reactions, both positive and negative, towards the works produced using AI. After reading various discussions and exchanges, I can summarize the points as follows:
Positive Reactions: There’s a much greater variety of artwork being produced, opening up opportunities for people without a background in art, writing, or storytelling to create and share their work with a wider audience. This lowers the “barrier to entry.”
Negative Reactions: Mostly, it’s the artists who’ve invested considerable time and effort in their craft—be it writing or painting—putting in intense concentration, utilizing mindfulness, pouring their hearts and souls, and providing substantial proof of work to create something truly unique to their style.
Some people view AI-generated works as lacking detail, careness, or vitality. They go as far as labeling them as “trash” or “rubbish.” When well-known artists express their strong feelings, it sparks intense debates, with both supporters and detractors expressing their opinions fervently. However, when you break it down, it all boils down to the positive and negative reactions I’ve summarized here.
It’s like deciding how much gasoline to pour on a perfectly ripe BBQ fire to turn the field into flames. It’s up to each individual to decide how far they want to fan the flames of discussion, whether it’s agreeing or disagreeing.
So, what do I think about all this?
Embarking on the AI Artist Journey in Anime-style Illustrations: I’ve never been particularly good at art; my drawing skills are abysmal. Even kindergarten kids like my 7 years old boy can draw better than me. During my final year of high school, when I needed to boost my grades to get into university, art was the subject that consistently dragged my GPA down. I think my art teachers could see right through me, so they let me pass just to avoid having to retake the exam. Meanwhile, my friends who were born for the arts could talk to the art teachers in the same language.
While my mind brims with ideas, I’ve struggled to materialize them. However, AI image generation has been an incredibly helpful tool in bringing my ideas to life. Sometimes, it exceeds my expectations by far.
Of course, the images generated by AI aren’t perfect from the get-go. They require refinement, detail tweaking, and sometimes, the assistance of photo editing software. This is where I’ve experienced a steep learning curve. It started off slow, but with practice, the time spent has decreased significantly, to the point where it’s become almost second nature (muscle memory).
Venturing into Writing: I took the leap into writing my first story on Rightshift after studying the history of global finance, delving into the deep-seated roots of problems embedded within the world’s financial system. When I reached a certain point in my studies, I wanted to share these stories with others through my writing style. But there were moments when I didn’t know where to find information or how to organize it cohesively. I turned to chatGPT and perplexity AI for help. Eventually, I took the suggestions from my personal assistant on the screen and incorporated them to align with my own style.
I See Many People Are Overly Fearful: “AI will take people’s jobs.” “Many jobs will be lost when AI becomes more prevalent.”
However, I don’t agree with the wave of fear that’s sweeping through discussions.
What if I told you: “AI isn’t here to take jobs away from people; it’s here to take jobs away from those who don’t know how to use AI.”
The truth is, AI’s accessibility means it opens up opportunities for people to learn, grow, and use it freely. While there may be some limitations in certain complex tasks but for the average user, it’s more than sufficient.
Take my experience with chatGPT, for instance. I used its suggestions to reduce the mortality rate of hospitalized patients or to shorten the last paragraph of a Minimalism article. I didn’t do anything extraordinary; I simply asked and asked, then adjusted the suggestions to fit my style and created the output. I didn’t blindly copy-paste everything AI suggested. If I had done that, it would have been pure garbage, as some artists said.
Even in image generation, within the Facebook group I joined for learning, prominent AI artists share profound ideas about generating images. I can’t recall their exact words, but one of them said something like:
The images generated aren’t considered fine artworks yet. Because it’s just something simple that anyone can generate by typing in a prompt. But uniqueness after using image editing softwares.
And that’s exactly it, because the principle of generating images is that we have to choose a model that fits the style of work we want. Some models are trained to create realistic images of people, while others are trained to create cartoon images in anime style. The process of training these models involves feeding them with a large dataset of hundreds or thousands of images as input for the AI to learn from. So when entering prompts, it produces ones that resemble the original ones it was trained on.
I’m lucky to come across this message, which made me realize that simply inputting a prompt and generating an image doesn’t make me an AI artist. It becomes a compelling point for me to delve into using image editing programs.
Generating images initially doesn’t take long. You just hit generate, leave the computer, take a shower, and come back to see. But when it comes to editing and refining the details of the work, that’s where the real time is spent, “the whole night.”
As I’ve honed my skills using this tool, the process of writing articles has become much shorter. I generate my own illustrations and then edit them. It doesn’t take long to finish. Back in the day, I had to search Google for illustrations, and often encountered problems like “not finding a suitable image” or “finding an image I like but it’s copyrighted.”
Writing articles doesn’t take long, but finding illustrations takes longer.
Can you see the picture where AI is a co-pilot sometimes when I create contents?
That’s exactly it, I agree with this statement:
“AI isn’t taking away jobs from humans; it’s taking away jobs from those who don’t know how to use AI.”
When I consider the capabilities of AI, it reminds me of situations where people have to adapt, and eventually there will be three groups of people:
The first group are those who study, try, and understand it well.
The second group are those who study, try, and understand it to some extent, but may not be as proficient as those in the first group.
The last group are those who refuse to learn or understand it.
It’s analogous to the various perspectives on Bitcoin. I’ve written about it in an article titled “Learn bitcoin before you have to; Study Bitcoin before necessity constrains or kills you” (English version also coming soon).
What I want to share with Everyone Reading Until the End
As of the day I’m writing this article, it’s April 27, 2024. New technologies are emerging so rapidly that it’s challenging to keep up. I firmly believe that there must be some tools that each of us can harness, tools that awaken certain potentials within us, even if we ourselves may not be fully aware of these capabilities. If we diligently practice and refine ourselves, these tools can accelerate our potential, unlock certain barriers, and tasks that once took a long time to complete may now be accomplished in just a few minutes or even seconds.
However, it’s important to acknowledge that we cannot keep up with everything. Attempting to do so can lead to exhaustion and disrupt life balance, causing us to lose relationships unnecessarily. Ultimately, I’d like to share the words of wisdom from the headmaster of the international school where my son studies:
“You should know something about everything, but you must know everything about something.” Until next time , Take care
#ai #storyteller #storytelling #article #articles