Life can be pretty scary for creators right now. The rise of AI language models like ChatGPT that can produce somewhat convincing pieces of writing, as well as the growing popularity of AI art, have all seen creatives re-evaluate their careers. After all, why would someone pay through the nose for a carefully crafted, curated piece of work, when the sophistication of this emerging technology is progressing by leaps and bounds with no sign of stopping or slowing down?
Photo: Digital Information World - AIgen
One of the most seemingly-vulnerable occupations in the line of AI’s fire is Journalism. The vocation to deliver news and current events to the people seems to be perfectly tailored to the efficiency and ease of use of AI. After all, where once you may have had to get an on-campus or online journalism degree such as a master's, now we can simply construct an AI program with the necessary crafting process, feed it enough content so that it learns the tone and structure of what we want from it, then feed it the data we want it to write about. Boom, journalism.
Right?
Let’s not think of AI as a program for a moment. Let’s think of it as a brain. We already know computers can accept input and produce output - for example, pressing a key (input) to print a letter to the screen (output). This is a process of receiving, understanding, and responding to stimuli, just like a human brain.
Image: Stefano Bucciarelli/Unsplash
When an AI program is first written, it’s just like the brain in a newborn baby. Inexperienced, curious, and imprintable. AI models begin with a “supervised learning” period, where the creators feed the AI brain a whole bunch of input that is defined. In the context of a baby, this can be likened to babies learning “mama” and “dada” as their first words; through continued exposure, the infant learns the names of its caregivers. The more data that it is exposed to, the more the baby learns, retains, and can accurately respond to. After a while, the AI will have acquired enough knowledge to undergo unsupervised learning, where after being given some parameters, the AI is then allowed to go through unlabelled data and learn what it could. In the context of a child, we can see this as going to daycare, and eventually school.
The next step is called the “Transformer Architecture,” and it performs a task that we often think of as unique to human brains - it draws from established knowledge to reach contextual conclusions about new knowledge. For example, if a child has only ever seen and used chairs before, they will likely be able to look at a barstool for the first time and, using their understanding of what chairs look like, where they’re used, and how they function, will be able to ascertain what a barstool is and how to use it. The transformer architecture does the same thing.
There is certainly more to it, but that’s the basic process. A program is created and fed defined data, then it’s fed undefined data, that it then defines itself, and then it uses the transformer architecture to contextually define more data. This is how ChatGPT and other AI models of its kind can absorb, understand, and then accurately respond to questions and prompts provided by the users.
Photo by Tyler Franta on Unsplash
Now, we stand at the precipice of a new industrial revolution. Although it can be easy to look at the things ChatGPT does and believe that it will dominate, possibly even wipe out career journalism and other industries, it appears to be more of a panic reaction than a logical one.
Although AI will inevitably disrupt industries, industry disruption likewise brings additional opportunity and scope. Not only that, but AI models like ChatGPT are capable of summarizing studies and articles, producing more efficient data for research in journalistic and expository writing pieces.
To illustrate the difference, allow us to put it in the following terms.
An AI will be able to produce a passable script on the effects of war by using data, statistics, eyewitness accounts, images, and videos. However, an AI cannot go to the scene of the war, develop its unique impressions, and produce a written or video piece with the creative decisions of a human journalist. An AI-created piece on the horrors of war would resemble something more akin to a documentary where a person in a seat just sits and rattles off what happened. You will learn, but you won’t understand, feel, or be stirred by it.
Modern journalism is just as much art as it is fact, and although this can produce some discrepancy and ethical dilemmas where one overtakes the other, it is the hints of humanity that make journalism a timeless profession. People will forever need to know what is going on in the world, but it’s the actions of journalists and the people they report on that put these events into a more “human” context; such as that people aren’t just aware of these events, they are moved by them.
If you’re a journalist concerned about your job once AI hits the main stage, we have some encouraging words for you. First of all, AI has already hit the main stage and you’re still here. Second, an AI can never do what you do. Finally, maybe you will be the person behind the next big development of the collaboration between AI and journalists. Either way, AI is merely a hammer, and that’s all it’s going to be for a long time. It’s up to you to drive the nail in.
One of the most seemingly-vulnerable occupations in the line of AI’s fire is Journalism. The vocation to deliver news and current events to the people seems to be perfectly tailored to the efficiency and ease of use of AI. After all, where once you may have had to get an on-campus or online journalism degree such as a master's, now we can simply construct an AI program with the necessary crafting process, feed it enough content so that it learns the tone and structure of what we want from it, then feed it the data we want it to write about. Boom, journalism.
Right?
How AI Works
Artificial Intelligence has been around for ages. Every time you use a search engine, you’re using an AI. GPS programs and devices use AI to extrapolate the best routes to a destination with consideration of traffic and roadworks. Although integrated into so many aspects of our lives, even healthcare, very few people understand how AI works, and how it arrives at the conclusions that it does.Let’s not think of AI as a program for a moment. Let’s think of it as a brain. We already know computers can accept input and produce output - for example, pressing a key (input) to print a letter to the screen (output). This is a process of receiving, understanding, and responding to stimuli, just like a human brain.
When an AI program is first written, it’s just like the brain in a newborn baby. Inexperienced, curious, and imprintable. AI models begin with a “supervised learning” period, where the creators feed the AI brain a whole bunch of input that is defined. In the context of a baby, this can be likened to babies learning “mama” and “dada” as their first words; through continued exposure, the infant learns the names of its caregivers. The more data that it is exposed to, the more the baby learns, retains, and can accurately respond to. After a while, the AI will have acquired enough knowledge to undergo unsupervised learning, where after being given some parameters, the AI is then allowed to go through unlabelled data and learn what it could. In the context of a child, we can see this as going to daycare, and eventually school.
The next step is called the “Transformer Architecture,” and it performs a task that we often think of as unique to human brains - it draws from established knowledge to reach contextual conclusions about new knowledge. For example, if a child has only ever seen and used chairs before, they will likely be able to look at a barstool for the first time and, using their understanding of what chairs look like, where they’re used, and how they function, will be able to ascertain what a barstool is and how to use it. The transformer architecture does the same thing.
There is certainly more to it, but that’s the basic process. A program is created and fed defined data, then it’s fed undefined data, that it then defines itself, and then it uses the transformer architecture to contextually define more data. This is how ChatGPT and other AI models of its kind can absorb, understand, and then accurately respond to questions and prompts provided by the users.
What This Means for Journalists and Journalism
Technology is constantly being developed, and there are always new ways to research and present information. Over the years there has been a dramatic shift in traditional media being responsible for relaying accurate, trustworthy information to the public, to online news outlets and content creators being the dominant source of information regarding current events.Now, we stand at the precipice of a new industrial revolution. Although it can be easy to look at the things ChatGPT does and believe that it will dominate, possibly even wipe out career journalism and other industries, it appears to be more of a panic reaction than a logical one.
Although AI will inevitably disrupt industries, industry disruption likewise brings additional opportunity and scope. Not only that, but AI models like ChatGPT are capable of summarizing studies and articles, producing more efficient data for research in journalistic and expository writing pieces.
To illustrate the difference, allow us to put it in the following terms.
An AI will be able to produce a passable script on the effects of war by using data, statistics, eyewitness accounts, images, and videos. However, an AI cannot go to the scene of the war, develop its unique impressions, and produce a written or video piece with the creative decisions of a human journalist. An AI-created piece on the horrors of war would resemble something more akin to a documentary where a person in a seat just sits and rattles off what happened. You will learn, but you won’t understand, feel, or be stirred by it.
Modern journalism is just as much art as it is fact, and although this can produce some discrepancy and ethical dilemmas where one overtakes the other, it is the hints of humanity that make journalism a timeless profession. People will forever need to know what is going on in the world, but it’s the actions of journalists and the people they report on that put these events into a more “human” context; such as that people aren’t just aware of these events, they are moved by them.
If you’re a journalist concerned about your job once AI hits the main stage, we have some encouraging words for you. First of all, AI has already hit the main stage and you’re still here. Second, an AI can never do what you do. Finally, maybe you will be the person behind the next big development of the collaboration between AI and journalists. Either way, AI is merely a hammer, and that’s all it’s going to be for a long time. It’s up to you to drive the nail in.
