Natural Language Generators
Could a Robot write this for me?
If I were to ask you whether what you’re reading right now was written by a human or a machine, you should feel confident in answering human. You would probably be right as well. But what if I asked you that same question in 15 years? Your answer should surprise you. In 15 years 90% of the news you read could be autonomously written.
The mind behind this prediction, Kristian Hammond, serves as CTO for Narrative Science. They developed and sell Quill, their unique automated word processor. Quill promises to analyze your data, write your business reports, produce investment portfolios and other tasks involving language and communication skills, basically all the boring, important part of work. It does this through a process of advanced natural language generation (NLG) which Narrative Science claims can ‘transform data into narrative’. This claim has so far been successful, as Quill now churns out lucid and detailed reports for 70 firms, mostly large and established financial institutions that make use of Quill to sift and report on more data than any human employer could hope to do.
Another company, Automated Insights, offers a similar product with Wordsmith. The claims it makes are similar, ‘Wordsmith makes writing thousands of articles as easy as writing one’. Although Wordsmith at least offers some degree of personalization in what it produces, both these companies have begun a process of automating human written interaction. The fact that machines can now apparently write as well as humans is both disappointing and also understandable. Grammarly, a start-up that offers grammar corrections based on 150 rules of English grammar, began after it noted that top organizations spend around $3 billion annually to improve employee language skills to a baseline level. That’s not even good English, but a baseline level that at least makes some sense.
This decline in the English language has been noticed by employers worldwide. It is most likely due to our information society which will abbreviate already short words and has zero consideration for basic English grammar. Quill and Wordsmith are therefore miracles to employers, they can finally get rid of all the time-consuming, poorly worded reports and allow their employees to focus on what they’re good at (presumably not writing reports), be it designing new products or strategizing responses to market changes. They can bring in countless benefits to financial institutions and small start-ups whose time is too valuable to be spent analyzing reams of data. Quill can even cut out marketing teams by producing effective product descriptions given information about a product.
Yet, despite the benefits, there is an almost existential threat posed by advanced NLG. There has been a consistent development of tech innovations or advances that can replace humans, but most of those are jobs where there are set instructions as to how to do things. Writing is different in that it is an inherently human and individual thing. What does it mean to us that machines can now write better than us? Aside from the fact that my work should be substantially easier in the future, just feed data into a program, it does mean a loss of traditionally human values, writing being a uniquely human and individual action.
This is not yet necessarily the case, although it could progress to that stage. As they are now, advanced NLGs are extremely useful to commercial firms where data is regularly generated by the markets they operate in. Such data would be readily available for analysis and subsequent communication in the form of reports by these NLGs. However, they appear to have underestimated the amount of work required in collecting information in other fields such as news, where data is not readily available and most of it is in a visual or verbal medium. Their compatibility with Tileo, a software that converts data into image reports such as graphs suggests a move towards the visual medium, but it is as yet limited to data extrapolation. So far, the technology appears promising, and set to streamline and increase efficiency in commercial and financial institutions. It still has a way to go in replacing writers.
Andre Hirschler 