Artificial Intelligence:

Can AI replace human writers?

September 20. 2020

Courtesy: Invento Robotics


Today AI can almost write like a humans do. But more advances in this technology are being made. One may ask how could an AI system learn to write as well, or say better than a human?



Well, in theory, one needs to continuously predict the next word based on what has already been written. It may sound a bit perplexing but if you have been using Gmail, you can see AI in action every day-- especially when the system suggests words so that you can finish your sentences. Or, have you noticed while replying back to text messages, you may see suggested responses on your mobile phone. You may not care as a consumer that it’s AI. You sure do appreciate the convenience it provides.



And this is just the beginning. Today there is a race to train AI systems to generate human language at scale. When this is achieved, the implications, both good or bad, are immense. Here are some examples of efficient writers for the time being.


GPT-3, The Robo-writer:


OpenAI, is a non-profit AI laboratory based in San Francisco backed by the likes of Elon Musk, Peter Thiel and Reid Hoffman, and it builds AI models. This laboratory has recently developed a Generative Pre-Trained Transformer (GPT) in 2018 and the following year GPT-2 arrived. The technology is an AI language generation model that automatically produces human-sounding language at scale. GPT-2 has the ability to construct long form content in different styles, using huge amounts of content drawn from the internet. Critics claim that the GPT-2 model had such substantial implications for malicious use that OpenAI originally chose not to release the trained model.  The laboratory had hoped that by limiting the release they will give the AI community enough time to ponder over the larger implications of such systems. But then in May 2020, OpenAI introduced a dramatically more powerful model called (predictably) GPT-3, a new text-generating program that can digest large swaths of the web, including coding tutorials.



100X data:


GPT-3 uses 100x as much data as its predecessor and now has begun to display incredible content creation capabilities. This includes turning text into code, evaluating investment memos and it is also capable to generate memes, poems, tweets, and even guitar tabs.



Limitation of GPT-3:


Though the special AI software can mimic human writing using the new language tool, but it falls short on common sense and a feeling brain. This technology raises major opportunities and challenges for marketers today. Some experts think an emerging technique called neuro-symbolic AI may be the answer. Here tones and styles can be programmed into computer algorithms, but for now, this effort will take some efforts and financial resources.


GTP-3 and the PhDs:


Delian Asparouhov, an investor with Founders Fund, an early backer of Facebook and SpaceX cofounded by Peter Thiel, blogged that GPT-3 “provides 10,000 PhDs that are willing to converse with you.” Asparouhov fed GPT-3 the start of a memo on a prospective health care investment. The system added discussion of regulatory hurdles and wrote, “I will be comfortable with that risk, because of the massive upside and massive costs savings to the system.”


NAO ROBOT WRITER. Courtesy: IRobot 4Ever

The ‘polite’ technology:


Denver entrepreneur Elliot Turner found that GPT-3 can rephrase rude comments into polite ones—or vice versa to insert insults. An independent researcher known as Gwern Branwen generated a trove of literary GPT-3 content, including pastiches of Harry Potter in the styles of Ernest Hemingway and Jane Austen.

Read GPT-3 written article from Guardian:



Robots writing in news organizations:


According to Search Engine Journal, robot writers are already working on data-driven short reporting articles, like stock prices, quarterly earnings, sports games outcomes, and Olympic medal awards. The Washington Post, for instance, created 850 articles in 2017 with Heliograf bot, and Quill AI robot has the capability of collating data on thousands of games. It can produce thousands of articles instantaneously, replacing professional sportswriters.


The platform used by The Associated Press is Wordsmith by Automated Insights, which is a type of natural language generation (NLG) machine learning that can transform raw data into thousands of publishable stories. This robot can write 12 times the amount of humans.



Will Robot Writers overtake humans?


Human writers are an amalgamation of life, philosophical, cultural, and intuitive ideas and experiences. We have empathy, which is a trait that's reserved for living creatures, particularly humans and other high-level primates and mammals. But can AI replicate thee traits. Though this will take a while but these questions are vehemently being explored. A recent study by Karstad University in Sweden found out that most people couldn't distinguish between news reports written by journalists and those generated by robo-writers.


If that does not surprise you consider the following advances…


**We'll have another decade or two to reach a point where robo-writers could dominate the publishing world.


**The potential for AI writers to reach "near perfection," in terms of their works will look "almost identical" to that of human writers, is high.


**An area that robo-writers are not likely to master is intuitive and empathetic writing.


**Recently in Japan an entire novel was written by an AI writer, and it also passed the first round of screening for the Hoshi Shinichi Literary Award in 2016


** Natural Language Generation is still limited to data feeding, so the algorithms can interpret and write a narrative about it.


**Human writers are still required to create templates and guidelines for those bots to follow


** AI and robo-writers lack emotions so we're obliged to work hard and have to be smart in optimizing our human instincts as individuals and writers.



What happens to content creators then?


**AI may create more jobs than it makes obsolete in the marketing industry at large.


**Professionals may have to re-evaluate their roles and skills, should AI technology become widely commercially available.


**Newspaper and magazine journalists were the among the first ones to be hit by the Internet. Journalists today have morphed themselves into content creators.


**No longer do they work for conventional publishers, but they work for media companies, including brands.