There are many trends in professional communication that can change the field. However, artificial intelligence that can write is the biggest game changer on the horizon. One application of AI writers is code documentation–a genre of documents that describes what users see on a webpage or software program.
I’m going to build my simple example to project what an AI writer might be capable of. I will base the example off of HTML because I can access a simple example of that code. However, my argument still applies to any other GUI interface with which users interact.
About Documenting Web Pages
HTML Developers are the ones who work under the hood with the real face of the internet. Rather than the visual interfaces and smooth navigation that so many people think of as the internet, developers understand that the internet is really millions of lines of code. Each line of code is an instruction for a browser page to behave in a certain way. All the lines together make the banners, images, headings, layout, and content the way a person sees them on the screen.
While developers do a lot of documentation, Technical Writers often describe their company’s websites in documentation so developers can focus on coding. Developers use that documentation to come back later, after writing or editing the code of other sites, and pick up the code where they left off. Alternatively, new developers are often assigned to the older web projects and pick up the code where the previous developers left off. The documentation describes each asset, function, feature, and behavior a user can see or click on a browser page.
In a way, all software documentation is like the methodology of a science experiment; a scientist carefully records the methodology so another scientist can recreate the experiment, with the same objectives, motives, and expectations. Likewise, a developer can rely on documentation to rebuild a website or software program from scratch–exactly like the original designer intended.
Jason Lawrence about the Future of Artificial Intelligence and Documentation
Once the code is completed, a user can use the website and utilize an interface to reach user goals. If the code already describes the site accurately and can present the site visually, then why can’t an AI describe the site textually? Why can’t an AI write the documentation? If documenting the code is simply a description of assets, functions, features, and behaviors then the AI can do that.
In fact, coding conventions label the assets already so the code contains information needed for documentation. For instance, <tb> is the code for a table. AI writers wouldn’t need to comprehend the concept of a table in order to spell out “table” in the documentation every time <tb> is mentioned in the code. Even further, <Div> codes let developers directly label clusters of assets by the name of the feature or function. AI writers would know that a <div id “banner”> is actually the banner across the top of the screen. An AI wouldn’t need to comprehend what a banner is in order to write about it because the nouns, verbs, and subjects are already there in the code.
Smiley Faces on the Web
I can illustrate this with a simple example. I can draw a smiley face on paper, scan it, and post the picture online. Alternatively, I can use a paint program to draw the smiley face and post the picture online. In either case, I can write a technical description of that smiley face so, detail its function, and its importance in the documentation.
The smiley face picture below is different. The smiley face is not drawn. This smiley face was not created either by pen or software. This smiley face was created by HTML code. The code details the positioning of colored circles.
I can write a technical description of this asset–this smiley face. Insofar as the positioning of the text caption and colored circles are all determined by code, an AI could potentially write that technical description as well.
Documenting the Code
If a browser program can present these lines of code visually, then why can’t an AI writer describe it? What is so complicated about a smiley face that only human intelligence can possible hope to describe it? Is it really so far beyond the capacity for an AI to describe a smiley face, seeing how all the information to describe the face is all right there in the code? The below code uses pixel coordinates to place specifically sized shapes. It uses simple color codes to identify which assets are yellow or black. Everything you need for smiley face is right there.
The only thing missing is the developer reflection. However, code comments are always embedded next to the respective elements. <!–comments–> can be placed anywhere in the HTML code and the browser knows to ignore anything within the tag. However, an AI would know to look for comments; the AI would know to incorporate code comments in the document’s description of the smiley face, website, or software application.
The Future
AI can already write. There are already successful AI software experiments and even a growing market for some AI writers. For instance, sports news off the wire is often written by AI. AI is writing finance reports, business intelligence reports, and scheduling appointments. AI writers will be in the workplace alongside human writers is near future. The code documentation may be one area in which AI writers are useful.
Jason Lawrence, M.S., Ph.D.
Driving the Conversation on Artificial Intelligence
