A few days ago, I had performed a Turing test on one of my best critics and keen readers so far: my wife. I showed her a poem and asked her what she thought of it. “It’s a bit repetitive, but otherwise the usual; you have much better pieces,” she had remarked. When I told her that it was in fact written by two authors, one human and one a non-human entity, she was deeply puzzled. The poem goes something like this:
“Before you knew
I spoke to myself,
and I knew
I had known you
before the universe was born.
I’m thinking, the sun’s going to get
down, so I’ll just put it this way,
maybe it’s going to rain
because I’ve got you and you don’t want it to rain,
you’re all that’s left
of me and I’m dying.”
If you cannot tell which lines were written by me or my co-author AI (Artificial Intelligence), then this was probably well worth your read.
With a very low cost in computational resources, this would mean that content creators with limited or no funds could take advantage of this in a very efficient fashion.
Before this, I had stumbled upon an organisation called Open AI which had in February unveiled “a language model called GPT-2 that generates coherent paragraphs of text one word at a time.”
According to their website, Talk to Transformer, the organisation, has decided only to release “small and medium-sized versions of it which aren’t as coherent but still produce interesting results” such as the above one.
When I first tested the bot, I was awestruck by the sheer amount of possibilities this technology could lead to. The technology that has been producing these kinds of texts is called machine learning.
Machine learning is a widely used application of AI that allows programmes to learn from extensive datasets without being programmed manually. At the heart of this application are the Artificial Neural Networks, which mimic human neurones and use pattern recognition on a broad variety of data categories and sets to learn and produce similar and fascinating results by themselves.
Uses of this kind of text generation are already happening at large, without many of us understanding it. Take for example the fact that it has the potential to provide an additional tool for content creation. With a very low cost in computational resources, this would mean that content creators with limited or no funds could take advantage of this in a very efficient fashion. Such content, that is, could replace advertisements with texts and be used as a substitute for, say, ads placed by internet businesses or other third parties, thereby creating a more efficient alternative to such marketing techniques.
The words after “the fact that it has” in the previous paragraph were generated by AI. As you can see, this is an extremely efficient way to cut budget for writers—both creative and non-creative. Moreover, it can replace, as the paragraph itself implies, certain writing tasks being automated, leading to job loss for low-cost/low-skilled writers.
Most of us have come across the debate as to whether AI will replace us. Some experts say that they will, sooner or later, but others have been on the sceptical side.
I have always been a firm believer in human creativity and originality. But seeing new inventions like these sometimes casts doubt on such beliefs. Being creative is a process, not everyone is born creative; it is for the most part a process of rigorous training of experimenting and coming up with aesthetic ideas and innovation. This process involves studying the works of the older generation and then creating something new out of it. If this is an accepted definition of creativity and the creative process, then we are at danger here. Imagine this: it takes almost half a lifetime for a human being to read enough to be able to pick up the art of writing and then actually write and get published, let alone be exceptionally adept in it. It takes a long time to study and absorb the amount of creativity or skills demonstrated above. Imagine a machine doing the same thing—only at a faster speed. What would it mean for the new generation of human writers?
And then, there are concerns for the readers as well. We are already immersed in a world of text and information. Human labour has value, and that is why we still patronise such labour. If you cannot differentiate the text written by a human author from that written by a machine, would you be willing to pay for it as much as you did before?
What about trust and validity? Imagine a rogue programmer, creating a programme to spread fake news. AI is not only good at generating text, but also visuals. Coupled with these, and the wretched wisdom of the crowd, a potentially harmful and seemingly innocent fake piece of news could easily bombard everyone’s screen, thus instigating acts of mass violence and hatred. In fact, this may already be happening, and maybe we are not fully aware of it yet.
Does this mean the death of the author as we know it? Does this mean that the profession of writing will eventually perish? I think not. At least not yet. Human creativity, apart from following others and learning certain strategies, also requires raw feelings and emotions. This is the most essential element of being human. We can turn our emotions into a work of art whereas machines can only mimic what has already been felt by others and reproduce from it.
So, what lies in the future?
I can only anticipate that the profession of writing will become a lot harder to pursue in the coming days. With that, the demand of human creativity will also rise. The machine cannot create anything on its own. It still needs a human guide. The only hope I see for the near future is collaboration between machines and human writers where, rather than competing with each other, both would complement each other’s skills and continue to produce great reads.
Muhammad Mustafa Monowar is currently studying Philosophy of Mind and Cognitive Science at the University of Birmingham.