Close ↓

AUTOMATION is impacting many industries including manufacturing, farming and transportation. But all these are blue-collar industries where it obviously makes sense to have a machine do the heavy lifting, so to speak. What about white collar work like writing, for instance?

Although Artificial Intelligence for the writing profession is still in its infancy, AI is set to play an increasingly important role in the years to come.

SMART COMPOSE

A very basic form of AI for writing is something you’ve probably already experienced but perhaps didn’t realise it. Smart Compose is a feature that Google introduced for Gmail last year. It basically suggests endings to sentences as you type them. I guess you could call it predictive text, which appears in grey. If it’s what you wanted to say, you just hit the tab button or if it’s not quite what you wanted to say, you can just ignore it. I’ve found it to be pretty accurate so most of the time I use it.

The concept of Smart Compose was inspired by writing computer code. Coding often requires long strings of identical sequences, so programmers use shortcuts to complete the code. Well, it so happens with human communication through e-mail, people tend to write the same things too.

For example, if you start the sentence with “Sorry to hear about your injury. Hope you get…” the ending is likely to be “well soon”. Instead of having to type “well soon”, Smart Compose does it for you and all you have to do is accept it.

All that is simple enough but how far can AI go in terms of writing articles? Open AI, an artificial intelligence company, has released an AI writer called GPT-2, which is best described as Smart Compose on steroids.

GPT-2 was trained to write from a 40 gigabytes worth of articles that people had posted links to on Reddit. In all, this involved about eight million webpages. Without human supervision, the neural net learned about the dynamics of language by analysing and computing the statistical probabilities of all the possible word combinations in this set of training data.

Its algorithm was designed so that once a human writer inputs a few sentences to establish the theme of the article, GPT-2 could then produce paragraphs of text to complete the story. This sounds scary but just how good is the system?

The BBC in August decided to take it for a test drive by inputting the first few sentences of a real story.

The first sentence (which was inputted) was: “Chinese search engine Sogou is creating artificial intelligence lookalikes to read popular novels in authors' voices.”

And this was the story that GPT-2 produced:

According to the researchers, this method of reading is not just the brainchild of Amazon Kindle users. Users of Android devices are also a notable market for fake reading. There are several theories to explain this.

Firstly, the device in a user's hand is different from the tablet or computer they are reading on. Secondly, the device used to access the Internet might be tampered with in some way. A third reason is the user may just copy and paste content directly into their virtual characters.

However, another explanation is that the device does not have voice recognition capabilities at all. If this is the case then voice recognition is more advanced than Google's translation engine. As such, these users are tricked into thinking they are reading from a real author – the book is made to sound authentic as soon as they try and read.

At a quick glance the sentences seem fine. The spelling and grammar are fine and the words seem to flow well enough. But if you read the sentences more carefully and try to make sense of them, you’ll realise that the article is gibberish. If you wish to compare it to the real article, go to https://bbc.in/33tPeqU.

As you can see, we’re still far from having machines replace authors and journalists. Still, like all technologies, this one will only improve with time. Will it eventually replace humans in the way that autonomous driving vehicles are expected to replace taxi drivers and truck drivers?

Personally, I doubt it. Writing is both an art and a science. The science part, which involves rules and structure, can be mastered by machines pretty easily but the art part is where the human touch comes in. That is much harder for machines to replicate.

For writers, instead of viewing AI with fear, they should view it as a tool to help them write better. Already, commercially, there are several tools that can be used to make your writing more efficient.

KEEPING IT CONCISE

Good writers know that the best writing is short, concise and to the point. Yet, even the best writers sometimes find themselves writing sentences that are too long and convoluted. Fortunately, there’s an app for that. The Hemingway App recommends ways to simplify your text. All you have to do is drop your article into the app and it will highlight problem areas in the text so you can work towards simplifying them.

PROOFREADING

Even something as common as Microsoft Word has an editor function that can check for spelling and basic grammatical mistakes. But if you want more powerful proofreading functionality, you can use Grammarly, which can help with contextual spelling, punctuation, grammar, style and even sentence structure.

READABILITY

Online tools like Atomic Reach are able to check the readability of your content. By using AI, it’s able to assess whether your sentences or paragraphs are too long, whether you have used the best transition words, whether your voice is too passive, and so on. According to the company, with just one click, your blog post can be optimised with “the ideal word choice, sentence structure, and more to help each segment form a connection with your brand”.

RESEARCH

To write a good article, you usually have to do a lot of research. The Internet is a treasure trove of information and sometimes it really feels like information overload. If you feel overwhelmed by research material, you can use Algorithmia’s Summariser, which helps provide smart summaries of long texts.

PLAGIARISM CHECK

Sometimes, writers inadvertently borrow too much from their research materials and could end up plagiarising other works. There are plenty of online tools, like Dupli Checker, to help check for plagiarism. Dupli Checker even has a Paraphrasing Tool to help you re-write a passage so that you don’t end up plagiarising it.

Oon Yeoh is a consultant with experiences in print, online and mobile media. Reach him at [email protected].

Close ↓