From publishers to news agencies to the business next door, many are already using AI to write. So, yes, AI-powered writing is here. This article explores some of the ways in which AI is enhancing a writer’s job now-a-days.
Before writing this article, I asked an AI system to help me write two essays. Check the results out here:
- Business Applications Of Artificial Intelligence
- Artificial Intelligence Everywhere
These are created by an AI from Articoolo. It is in ‘beta’ mode. Articoolo generates articles based on 2 to 5-word phrases that a user provides. Creates articles within a couple of minutes. Not a bad output for spending 2 minutes on each article. But, as you read them, you may realise that these are like essays by a primary school kid than an expert covering the topic.
Writing the first draft for an article is just one example. AI is helping writers in many more ways. For example:
- Interpret numbers/data into plain English. Narrative Science
- Get writing assistance beyond simple grammar and spelling checks through Microsoft Word, Grammarly and many more services.
- An AI called Heliograf writes hundreds of news articles on topics like Rio Olympics and US Presidential elections at The Washington Post. Press Association (UK), through a project called RADAR, uses AI to pen down 30,000+ local stories a month.
- Reuters’ News Tracer helps journalists spot and validate real news in real time on Twitter. Not only does it check if a tweet topic is newsworthy, but it also helps assess whether a piece of information is true.
- Spot statistical errors in research papers. This AI called Statcheck, tore apart 30,000 articles under 2 hours to find hundreds of statistical inconsistencies in published psychology papers.
- Peer review articles for publishers. Multiple solutions, like UNSILO, exist that help academic publishers screen initial papers through automated peer reviews by AI.
- Even write poems and generate poetry that is worth getting published.
Like the sample essays above, each of these AI systems have their limitations. They are at different levels of sophistication. However, they all have a common trait. They specialize in one narrow task, like poetry writing, and do it consistently. Experts classify these using the term Artificial Narrow Intelligence (ANI). Most practical and commercial applications of AI today belong to this type.
To be called intelligent, they need to get four basics right :
- process new pieces of information
- apply rules and/or logic related to the area
- interpret subjective/grey areas in those rules
- and communicate back with real people
Plus, they should ideally keep getting better with each iteration or correction.
Most of these AI systems also depend on the availability of tons of data to get trained. A few techniques are coming up to make AI systems learn and get better even with little data. We’ll discuss that in another post.