In November 2022, OpenAI, a tech company in the US, launched its ChatGPT AI tool. Within five days, ChatGPT had 1 million users. This was the quickest any software had reached a user base of a million. It wasn’t just that many people were using ChatGPT, but also that many more were talking about it. Over the next few weeks, social media was buzzing with what ChatGPT was capable of. If you haven’t heard of ChatGPT by now, I would be very surprised.
What is AI?
AI stands for artificial intelligence. In the past, we used to think that an AI would be a supercomputer with the ability to feel things and have human-like consciousness. However, creating a computer like that is impossible, so humanity has instead focused on creating Machine Learning AI.
What is Machine Learning?
Think about how a child grows up and learns things. At the beginning, he has no idea that a stove is hot and dangerous to touch. So, his curiosity gets the better of him and he touches the stove. His hand burns. He is in pain. So, he learns not to touch a stove anymore.
As he grows up a little more, he learns that the stove is actually safe to touch sometimes. By the time the child reaches his teenage years, he is aware of the nuances of when and where a stove is safe to touch. In other words, humans learn things through lived experience and by trial and error (of course, humans also learn things vicariously, i.e., through the live experience of others, but we’ll not talk about that for now).
This trial and error learning method is how you learn to distinguish a cat from a dog. At the start of your life, you didn’t know what a cat was. And then your parents introduced you to one. Then as you grew up, you saw more and more cats, so much so that you now know what a cat looks like from memory.
A Machine Learning AI is a computer that learns in a similar manner. Just like your parents taught you what a cat looks like, researchers teach computers to recognize cats in photos by making the computer look at millions of photos of cats. They can teach computers to know text like humans by showing millions of examples of human-human text interactions. The computers that can do all of these are what we now call AI.
What Are Some Examples of AI?
The most popular example of an AI is ChatGPT. It belongs to a model of AI called Large Language Models, which (as the “language” part of its name suggests) can generate human-like text. Go to Google, search for ChatGPT, and try it out for yourself today. If you want to write an email to someone but are worried that your English isn’t good, don’t worry about it. Just write that email in whatever English is natural to you, then go to ChatGPT and ask it to rewrite that email in proper English. You can even tell it to write the email in a specific tone (such as angry, respectful, confident, etc.) You can ask ChatGPT to write note sheets, essays, poems, anything.
Another powerful family of AI is Generative AI. As I mentioned earlier, researchers need to show computers millions of examples to teach them what a cat looks like. Similar to how humans can imagine what a cat looks like based on their memory, AI can generate pictures of cats. Text-to-Image AI Tools such as DALL-E, Midjourney, and Stable Diffusion take prompts from humans and generate images. If you go to DALL-E on your phone or computer and ask it to generate a happy yellow cat sitting on a carpet in the mountains of Bhutan, basking in the sun, it will give you exactly that image. These models promise that their generated images will be unique every time and that different users will receive different images even when given the same prompt.
If you want to make videos, generative AI tools such as Pictory AI can help. You can ask ChatGPT to write a script, then copy-paste that script into Pictory AI to create videos. If you are not happy with the sound of your video, tools such as Adobe’s Podcast can enhance voice recordings. This tool can clean wind sounds and make recordings sound like they were done in a studio.
There are several other examples of AI tools, and I hope to write about more of them someday. However, for now, we need to look at the bigger picture related to AI.
As part of my job at Nyingnor, I have provided training in the digital space to many people in Bhutan, including journalists from every media house. Recently, I started teaching AI tools, and the feedback has been overwhelmingly positive. However, one recent comment got me thinking. Someone said that these AI tools make them feel like we are outsourcing thinking to computers and that they worry about our future. This comment is not isolated, and there is growing sentiment around the world that AI tools are dangerous and could make humans dependent on computers.
All of this is valid, but I see it differently. Human history is full of comments like this. Socrates famously did not write things down because he thought the written word would make students lazy. When calculators were first invented, some people worried that they would make humans lose the ability to do their own math.
The history of the world and our lives is cumulative. Every technology is built on pre-existing ones, and sometimes, that means new technology makes older ones obsolete. However, that is more than compensated for by the collective increase in human productivity. Thanks to the written word, many more people learned maths, science, and philosophy in Ancient Greece. When Gutenberg introduced the printing press in Europe, knowledge became more easily accessible to more people than before. Every time a new technology is invented, the collective productivity of humanity goes up.
This should be our attitude towards AI. Of course, AI will make some existing jobs obsolete. As the CEO of a company that provides digital marketing services, I see that within the next few years, our graphic design services will see declining customer interest. However, we have to embrace AI and look for ways these tools can improve our efficiency.
An example of this is how ChatGPT, which can code in many languages, is being used by some senior engineers to write small portions of their code. In our office, thanks to Adobe’s Podcast, we now spend little to no time improving sound quality. Thanks to DALL-E and other tools, we spend less time on graphic design. Thanks to ChatGPT, we spend less time on research. Thanks to Tome AI, we spend less time making our presentations fanciful.
None of these tools mean that human input is redundant. It just means that we are more efficient than ever before. Since last year, I find that I am under less stress and that I have to work fewer hours than before. A part of this is thanks to improving my productivity and efficiency through the use of AI Tools.