Machine learning and artificial intelligence continue to develop and allow algorithms to write better and better, autonomously (books, articles, and all kinds of content). For the moment, they are very far from being real writers or authors, because their texts are limited to certain sectors where creativity, style, human judgment have little importance, such as economic results, sports results, weather, real estate announcements … But their progress is still significant, and it seems that we are getting closer every year to the moment when the authors will be replaced by machines. So where are we today? Are robot writers coming soon? What are the challenges, what is the future of autonomous writing managed by AI?
Automated writing is winning over the journalism industry, publishing and writing … and even fake reviews
AI writing seems to be advancing with leaps and bounds: while Microsoft recently decided to replace dozens of journalist positions on its MSN portal with algorithms, a new artificial intelligence from the American OpenAI, a start-up co-founded by the futuristic billionaire Elon Musk, was talked about in early September with a “model” called GPT-3 , which allows the user to start a sentence, and let the algorithms finish it. British newspaper The Guardian published on September 8, an editorial signed by GPT-3, with the ironic headline: “A robot wrote this whole article. Are you ever afraid, human? “(A robot wrote this entire article. Are you scared yet, human?)
The British daily actually entrusted the writing of this article to GPT-3 to demonstrate that humans had absolutely nothing to fear from robots, even if they are able to take charge of writing a perfectly coherent text. The Guardian explained that the tool produced eight versions of the text, and reporters selected the best passages to compose the published article. These snippets have been selected to show the different registers that the AI is capable of producing. Perhaps the most interesting point of this experiment is quite simply the demonstration that the work of writing assisted by algorithms is much faster than the 100% human writing.
The algorithm therefore rather offers an acceleration of the journalist’s work when it comes to a complex text, or a completely autonomous writing, for short texts such as announcements, advertisements, etc. Unfortunately, this type of tool can also be used for writing false opinion on e-commerce sites, or to mass produce articles for a disinformation campaign.
And the subjective in all this?
Although algorithms can be easily trained to collect and reformulate certain types of information, they struggle to understand exactly how to frame information in a subjectively appealing way.
While some articles are simply based on presenting an opinion or a fact, others require the development of a unique, original, subjective perspective, characteristic of human consciousness and experience. Algorithms are currently struggling to invent this type of new human thinking and opinion that seems realistic; they can only imitate what has already been presented.
AI to assist authors
Artificial intelligence is in reality not designed to replace humans, but to come to their aid either by putting the author’s words in writing thanks to automatic speech recognition, or by improving a created text by the author.
Happy scribe for example, is a solution based on voice recognition and artificial intelligence that transcribes any audio content in 119 languages and accents, in order to meet accessibility requirements for the deaf and hard of hearing, and SEO (referencing natural on search engines) or simply availability of words in written format.
End of August, Microsoft Word has also integrated a “transcription” function into its online version, which automatically transcribes an audio file. The taking of notes of interviews, interviews and meetings, audio files is thus done automatically, allowing a real saving of time.
For authors, the tool Marlow of the startup Authors IA corrects spelling and syntax errors and also offers alternatives to optimize texts. It analyzes the structure of the plot, the theme, the rhythm, the character of the characters and also works on the level of complexity of the book, the repetitive sentences, the dialogues and the narrative arc.
Today, algorithms are therefore allies in improving writing and making note-taking and transcription more agile. They are also quite capable of writing small texts for a sufficiently simple and easy to understand subject. However, in the near future, AI-based writing algorithms will still have shortcomings, based on their inability to replicate human emotions, their limited ability to deal with complex subjects, and their lack of “natural” creativity.