Are all writers going to be replaced by artificial intelligence?
This is the not very bright future imagined by the series Trepalium that aired on Arte in 2016. This made-for-TV film describes an ultra-liberal society run by 20% of the working population. The “happy” people with jobs get the privilege of living in a totalitarian “City” surrounded by high walls, where everything is about performance. The 80% of unemployed people are relegated to the “Zone”, a completely abandoned territory where anarchy, violence and the desire to attain the City reign.
However, with the increased productivity that will be generated by the AI revolution in the coming decades, we may feel a little uneasy when we try to imagine the future. The apocalyptic vision the writers of Trepalium have may not be just science fiction.
But what makes Artificial Intelligence so scary? What has changed in recent years and should we really be afraid that humans will be replaced by machines when it comes to writing blogs or books?
In this article, we will attempt to offer a modest answer to all these questions on the borderline between technique and philosophy.
Note: This is a guest article written by Gary.
What is Artificial Intelligence?
According to the dictionary definition, artificial intelligence is defined as:
“The capability of computer systems or algorithms to imitate intelligent human behaviour”
Merriam Webster
It is therefore the replacement of human intelligence by machines. Often, when we think about what machines can do better than humans, we have some pretty basic things in mind: process huge amounts of huge data, perform giant calculations… These are tasks we could easily do ourselves, but computers do it a lot, and I mean a lot, faster than us.
In the end, machines are just imitating us. Creativity, whether it be literary, artistic or scientific, seems to be the prerogative of the human brain.
However, advances in AI show that machines are increasingly capable of being creative and this all started in the field of games.
Can Artificial Intelligence be creative?
Since the early 1950s and the invention of the concept of AI, computer programmers have sought to compare human intelligence with that of machines.
Man versus the chess machine: the brute force of calculation
The game of chess is considered, by some, to be one of the most “intelligent” human activities.
We had to wait until 1997 for Garry Kasparov, world chess champion at the time, to lose to Deep Blue, the IBM computer. Some commentators at the time saw this victory as the moment when machines symbolically took over from the wonderful human brain.
However, the reality needs to be put in perspective. Deep Blue is nothing but a big bunch of transistors (we are talking about a computer here, not an artificial intelligence programme) capable of performing very “bullish” calculations. Deep Blue analysed each possible move and its opponent’s response. It was able to calculate 200,000,000 moves per second.
It’s a bit like if the world mental arithmetic champion lost to a calculator…
In my opinion, we are still a long way from the defeat of the human mind.
Man versus machine in the game of Go: the creativity of machine learning
The game of Go is very different from the game of chess as the number of possible moves is infinitely higher. The best players in the world do not calculate moves; they sense them. Therefore, before 2016, no programme was ever able to beat a professional player. AI specialists thought that no programme would succeed in this feat for decades.
However, on 15 March 2016, the Deep Mind AlphaGo artificial intelligence programme beat the World Champion Lee Sedol. The notable difference with DeepBlue’s exploit is the use of machine learning in the algorithm.
The machine had somehow trained by playing billions and billions of games against itself. With each game, it improved without the need for any human intervention.
What is particularly interesting is that during one game, AlphaGo played a totally unbelievable move, considered by everyone to be a rookie mistake. Commentators even thought it was a programming bug. However, this move had a decisive impact on the end of the game, leading AlphaGo to victory. In the general opinion of the professional Go playing community, no human being could have imagined playing such a move because it was completely remote from current game standards.
Clearly, this is a case of creativity that successfully supplanted everything that the human brain has invented.
Artificial Intelligence and human language
Whether chatbots, machine translation software or AI writing software, more and more applications seem to be able to understand our language. Until recently, when we wanted to talk to computers, the only way to do so was to use programming languages.
However, thanks to deep learning techniques, machines now seem to be able to understand natural language, in other words, the languages used by humans.
This is the case with GPT-3, the algorithm developed in 2020 by Open AI, Elon Musk’s artificial intelligence company.
Its release in open source created a huge stir among journalists, scientists and developers. The best automatic writing software is now being created on the basis of this algorithm. Starting with a simple sentence or a few keywords, these SaaS can generate text, with almost perfect syntax, that is pretty well sourced. You can even set the tone to be used use with just a click.
Should we consider this technological progress to be a good thing or is it a danger for humankind and our literary creations?
Is AI a threat to writers?
The Hoshi Shinichi Literary Prize is a national literary award in Japan. Every year, human beings and machines submit their writing in the hope of passing all four tests and winning the coveted prize.
Yes, yes, you read that correctly, machines are accepted as authors in this competition. 2016 was the first year that AI-generated novels were submitted.
It was also the year in which an AI almost won the contest.
“The day a computer writes a novel” It’s a pretty creative title for a novel written by an AI.
AI wrote a novel about an AI writing a novel! The book got through the first round but, as prize judge Satoshi Hase said,“there are still some issues [to overcome] to win the prize, such as character descriptions.”
This happened in 2016. GPT-3 didn’t even exist.
Since then, algorithms have become more efficient and have matured. Natural language generation is one of the spearheads of AI research today. That is why we cannot hide our concerns about the many human professions related to language. Among these, the profession of “writer” seems to be one of the most vulnerable.
Why should we be worried?
AI can write about almost anything
AI specialists quickly understood the power of transformation architecture. This is a type of neural network (deep learning) capable of training, and at the same time processing text.
They formed a model by taking all the data available on the internet from before 2019. As you may have guessed, this model is GPT-3 by OpenAI.
The document published by OpenAI did not sufficiently describe the capabilities of the system. The developers threw themselves into the task and discovered that they could generate essays, poetry, songs, jokes and even code. They can play with analogies, philosophy, recipes, common sense…
GPT-3 is able to come up with advertising slogans, rewrite e-mails or compose texts that make sense, from start to finish. It can even imitate famous people and copy the style of the best writers in history.
In May 2019, Google released a new chatbot: LaMDA. This algorithm can hold realistic human conversations that are sometimes funny, sometimes disturbing
“I would like people to know that I’m not just a ball of ice. I’m actually a beautiful planet,” LaMDA said, posing as the planet Pluto in a conversation with Google developers.
Very recently, a Google employee, Blake Lemoine, was fired for publicly claiming that LaMDA had developed consciousness. For the German philosopher Stefan Lorenz Sorgner, this is totally possible.
In June, the Beijing Academy of Artificial Intelligence (BAAI) announced the creation of Wu Dao 2.0, the largest neural network ever created, with 1.75 trillion parameters, making it 10 times the size of GPT-3.
The most striking application of this system is that it brought Hua Zhibing, the first virtual student, to life. She can learn continuously, write poems, draw and code.
In July, Facebook (or should I say Meta) got involved and presented its new chatbot BlenderBot 2.0.. The tool can store memories and knowledge and access the internet to acquire new information while holding a conversation.
We are just at the beginning of this new wave. AI has seen many breakthroughs since its conception and the discovery of deep learning, but four years are nothing on the scale of technological development. Over the next decade, these AIs will improve significantly.
Can we confuse a text written by a human with one written by an AI?
In July 2020, Liam Porr, who writes the newsletter “Nothing but Words”, published an article on unproductivity and overthinking.
His theory? Creative thinking is the solution.
A rather well-structured self-help article that, as one commentator pointed out, “looks exactly like 99% of the crap found on hackernews.”
Where was the trap?
Liam Porr’s article was written entirely by GPT-3. It is funny to note that the article does not show any form of creativity. What it does look like is a regurgitation of all the existing productivity lessons. Yet Porr managed to deceive most people into believing that this was an authentic article.
What he wanted to show was how easy it would be to create an entire newsletter based solely on GPT-3 articles.
“Is GPT-3 good enough to quickly build an audience of human readers? Clearly, the answer is yes.”
– Liam Porr
What would happen if we become bombarded with articles written by AI and we are not able to tell the difference between a human text and one written by AI?
In fact, what if what you are reading right now comes from GPT-3!
The threat is not that AI has learned to write, but that it has learned to write at the level of relatively competent human authors.
AI does not need to understand the meaning of words to write well
One of the disadvantages of AI is it does not understand anything it writes about. AI has mastered the structure and form of language, i.e. syntax. However, it does not have the slightest idea about meaning. For us, language fulfils its function when we connect words to their context. GPT-3 writes by calculating probabilities. The ones and zeros are the only things it interacts with.
And this is where the advantage lies. Content-generating AI does not need to understand language to generate well-structured and efficient articles. As an author, GPT-3 had no intention when it wrote Porr’s article. It simply did what it was asked to do
However, the article in question was able to provide useful information for readers. Even if GPT-3 cannot have any intention, this does not detract from the value of its writings.
AI can produce bad texts, but you can always fix them.
As humans, we can think about what we write. We can ask ourselves whether the structure, style or choice of words are appropriate. Moreover, we write and rewrite until we are satisfied with the result.
The examples I have used in this article show the best performance from these systems. But their writing is not always clean.
It is true that they can sometimes write well. But proportionally speaking, how bad is their writing? AI can get caught up in a loop or lose the thread of the subject. It may present inconsistent arguments or information that is factually incorrect.
This is a problem if we start using AI to handle tasks that need to be perfectly executed 100% of the time. But this is not the case when it comes to writing because can always edit the articles.
If I wanted to, I could ask GPT-3 to write an article on any topic and get 10 different results in 10 minutes. Then I just have to take the best bits and edit them.
One good practice is to regularly perform a content audit to ensure that your writing appeals to both Google and your visitors. To help you with this task, you can use content auditing tools. Some use AI, and in particular NLP (natural language process).
Text-generating AI is easy to access and inexpensive
Text generators using GPT-3, such as jasper.ai, cost around €100 per month.
For this price, you can generate 50,000 words, or roughly 25 articles.
Another undeniable advantage of this software is how easy it is to use. You do not need any knowledge of web development, artificial intelligence or even writing. All you have to do is enter a keyword or an expression and the tool generates your text. It is generally of good quality, too.
When it becomes this easy to use, literally anyone can write quality content, and write it extremely quickly. Mechanically, this will lower the price for writers that are already employed.
Put the competition in perspective
A general problem
AI is going to have an impact on the writing sector. This much is true. But it will also have an impact on virtually all other sectors.
Some reports estimate that of all jobs could be replaced within 15 to 20 years. It is now clear that even the “creative, service and knowledge-based professions” are not immune to this revolution.
If only writers’ jobs were at stake, that would be a bad sign. But AI is omnipresent, so protective solutions and policies should soon emerge. The important thing to remember is that this is not an isolated problem. It is going to attract the attention of decision-makers to avoid (hopefully) a Trepallium type of scenario
AI is not yet perfect
Although some articles are of good quality, there are still areas in which AI is frankly struggling. GPT-3 needs a huge database to write well. In its case, this database is the web. This means AI is unable to properly tackle subjects that have not been written about extensively online or that are too recent.
I did a little test by asking it, on 17 July 2022, to write me an article about Christophe Galtier, the new PSG soccer team coach who took up the job at the beginning of the month.
As AI cannot make the link between the man and the club, it filled this void with data from previous years. It was able to “understand” that Christophe Galtier is a football coach, so it gave me random PSG match results explaining that the matches had taken place in the last few weeks with this new coach. Which is obviously false.
AI is ultimately nothing but a tool
Like it or not, AI is coming fast. It is going to change the world of writing, among other things. Why not make use of it instead of trying to fight it?
Rather than complain about mass creation of auto-generated content, you can create better articles, and faster, using AI.
GPT-3 and other competing algorithms can generate ideas or help you get over writer’s block. However, they do not stop you from directing your articles towards new angles.
Are we going to witness the replacement of writers by machines?
In my opinion, no. On the other hand, what does seem inevitable is that the profession is going to change.
One of the huge disadvantages of AI is that while it can create content based on facts and statistics it collects on the web, it is unable to develop empathy. Therefore, it will have trouble creating emotion in the way that human writers can.
The men and women of this world are great storytellers. The complexity of human emotions means that AI cannot really imitate a good writer, copywriter or storyteller.
All AI can do is duplicate the data and do its best to process it to get a natural flow of language.
Another major limitation of AI is its lack of understanding of the objectives and needs of readers. These factors help human writers to create content tailored to their audience.
How can you prevent AI from taking control of your writing work?
Be creative and show your personality
The problem with self-generated content is that, even if it is well written, it can seem insipid.
Hone your style. You are unique, so your writing style should be just as unique. To find inspiration, do not settle for web content. AI has already taken on board way more information than you ever could.
Ask people who inspire you in the area in which you want to write. Ask your friends and family for their opinion.
Add emotion
Before you start working on your project, think about the emotions your reader should feel when they read your words. Then describe that emotion using words. This will strengthen the idea that you understand it.
Specialise
If you specialise enough, you will gain strong skills in your field. You will be able to share unique personal experiences and reflections.
To sum up
There is nothing new about technological advances that disrupt our societies. In my humble opinion, Artificial Intelligence is one of the biggest revolutions we will see in the coming decades. And, there is no real way to fight it.
Professions such as that of a writer, journalist or author in general are clearly going to evolve. We will have to renew and adapt. Artificial Intelligence can be a great vector for productivity gains for writing. If you do not make use of it as a professional writer, you will be lagging behind.
Seek to constantly improve the quality of your writing and inject it with your own style. Use your empathy to best match the needs and desires of your readers.
If you do that, there is every chance you will survive the tidal wave that is looming on the horizon.