ເນື້ອໃນການຜະລິດ

AI is coming for the written word, but does it have anything useful to say?

The term “industrial revolution” typically refers to a period of rapid economic and technological development that occurred in 18th and 19th century Europe. It was characterized by the widespread adoption of new technologies such as steam power, the mechanization of factories, and the development of new forms of transportation such as trains and ships.

In recent years, the concept of an “industrial revolution” has been applied to the rise of artificial intelligence (AI) and its potential to transform various industries. Just as the industrial revolution of the past was driven by new technologies, the AI revolution of the present is being driven by the rapid advancement of machine learning and other AI technologies.

AI is developing very fast, so fast that already it is nearly impossible to tell whether something was written by a computer. The previous two paragraphs were penned by OpenAI’s ສົນທະນາ GPT, an advanced chatbot that generates spookily human-like responses to almost any prompt. I asked it to compose an essay explaining how AI could lead to a new industrial revolution for ideas, and that’s just what it did. This isn’t a new trick, entire articles (even books) have been “written by AI”. Rather than use it as a gimmick, I want to take the computer’s response as a prompt, an opportunity to poke at the question of what AI means for the ways we generate and express ideas.

Part of what makes tools like ChatGPT so compelling is how intuitive it is to work with them. A common way of interacting is to speak, to ask a question or describe the image you wish to see; to literally “ask Jeeves”. It adds to the sense of engaging with a thinking being, but the AI isn’t ຕອບ a prompt, not really. Instead, it takes whatever is asked of it, and draws associations from a vast ocean of human writing on which it has been ‘trained’, collating the results into a form that resembles human writing. By reflecting the patterns it finds in what human beings have written. The result happens to be very, very convincing, but the machine doesn’t really ຮູ້ what it’s saying.

That’s not to say that the machines don’t have a certain character, or voice. Anyone who has read enough output from ChatGPT (or other AI systems) might have recognized it in the text at the beginning of this post. The machines are versatile, too, able to do everything from writing a song to correcting computer code, but it’s all still an expression of the original training data. It’s impressive, but it isn’t magic, and we should approach it as a tool rather than an oracle.

Where AI shines is in organizing information. In a sense, it recognizes the typical ‘shape’ of an argument or speech pattern (or objects and scenes, in the case of image-based systems like DALL-E). That means it’s not generating new insights, but reflecting our own ways of communicating and perceiving back to us, which in and of itself is amazing and full of potential. It’s like a mirror, one that can clarify or distort, and it’s sure to teach us a lot about ourselves.

To hear people talking about the implications of these new tools, you might think the role of the human brain is starting to shrink. But that isn’t so. At this point, and for the foreseeable future, these systems are limited by their training data, requiring input from human beings in order to generate anything novel. If you feed an AI content generated by an AI, the quality of the responses degrades. There has to be ‘blood in the machine’ for it to be generative, and create anything of interest or relevance to us humans. That might sound a little like Skynet, but it’s should be a comforting thought — AI isn’t taking over for humans, we’re giving it a reason to exist, and it can help us do what we do better, and faster.

After all, for anything an AI produces to be meaningful requires a human being at the other end to assign it meaning. I think a lot about where new and useful ideas come from. Ideas that get traction usually emerge from one or two minds, and then grow into more tangible forms, like new technologies, companies, or schools of thought. These accomplishments always build on the work of countless people that have come before, and the influence of countless others who create the conditions for a person to have an original thought and to explore it. AI has its greatest potential value where the vast network of human thought and discourse crystalize into a new idea. It may help to raise the standards for everybody, now able to do heavier mental ‘lifting’ than before. In this sense, it does seem a bit like a cognitive version of the Industrial Revolution. AI is ultimately a human augmentation device, not unlike the internet or the automobile.

In my own experience, ChatGPT has actually helped me come to some new insights. Not because the AI came up with an original thought, but because when I asked it to represent something I had been thinking about, it drew out connections that I might not have, which in turn inspired a new thought in my own mind. From this point, I can pose new prompts to the machine, and hopefully be spurred further. It is this process that epitomizes to me the potential of AI, as a means for human augmentation.

This leaves me optimistic that the transformative potential for AI tools is in helping us innovate and solve problems, and help us move toward a better world. Revolutionary advances are being made in generating new proteins and medicinal molecules, for instance, making high throughput testing more efficient, or just helping people who have trouble writing emails. Its power comes from being able to excel at the kinds of processing that humans struggle with, quickly teasing out patterns from gigantic amounts of data. I would love to feed an AI all of my team’s Slack messages, to parse threads for old ideas or new research insights.

What these machines won’t do is take over our thinking for us. They are closed loops, only able to reflect and restructure what has already been fed in. We can wield and shape those loops, peer into them to draw out insights, and improve the system to help us generate more new ideas, or more useful questions. Rather than worry about AI gaining consciousness or taking over the planet, we should be developing a working relationship with them as a means of achieving new insights and cracking hard problems.

The field is evolving at an astonishing rate, and much like today’s world would have been impossible to predict at the turn of the 20th century, there’s no telling what this will all lead to in the long run. No doubt a lot of amazing new AI based services are about to emerge, and I for one am excited to try them out. But since OpenAI hasn’t closed free access yet, we might as well ask the AI for its opinion:

Overall, the potential for AI to create an industrial revolution for ideas is significant. By automating routine tasks and generating new ideas, AI has the potential to drive innovation and improve efficiency in a wide range of industries. This could lead to a new era of economic growth and development, much like the industrial revolution of the past.