Mark TwAIn?: The expressive side of artificial intelligence
Trishan Chanda
“To write is human…”- Stephen King
As AI is steadily gaining a significant influence over our lives, the question must be asked as to whether there is anything it is unable to do. Albeit the obvious advantages that AI brings (i.e. its ability to complete tasks quickly and accurately), could there be a dimension to robots and artificial intelligence that does not conform to these expectations?
Isaac Asimov’s three laws of robotics have been pivotal to our understanding and design of robots to this day. But do they truly signify the limits of AI?
Imagine: it is the year 2060, and we have scraped our way past climate change to see another lifetime on planet Earth; a minimal chance, but for the sake of this article, we will assume so. Our automated counterparts have reached a point of domination in society because they have managed to triumph in the anticipated battle of human vs robot. You also have much more time in your hands given that your personalised robot does almost all your work for you, so you want to read a book. Perhaps running back the years with some J.K. Rowling? Not a chance. Time to look ahead with GPT-2.
GPT-2 is a transformer-based language model, a brainchild of Elon Musk’s tech company- OpenAI. It has the ability to write stories, articles and news reports, while it has been particularly recognised for its chameleon-like style of writing (i.e. its excellent impressions of various authors in order to make stories as realistic as possible). However, according to Elon Musk and his fellow entrepreneurs, this technology would be “too dangerous for them to release in the wild”. But why so?
Amidst a generation that is becoming more reliant on technology than ever, there is a growing controversy over the reliability of various sources, and hence this statement from the owners. Open AI fears that this new branch of technology could be exploited severely, to the extent where it may also deep fake news reports and articles, if found in the wrong hands. However, this proposition also has its drawbacks; is this not a bottomless claim when people are already creating fake news so effortlessly? If TikTok has the ability to convince some that Stranger Things is actually inspired from real-life discoveries, surely we have reached a point of rock bottom already?. Would it not make much more sense to use this technology in a much more positive manner and to develop this AI into a means of mitigating this significant issue?
To give an example of GPT-2’s functionality, the program was instructed to write a follow-up to the first lines of George Orwell’s 1984: “It was a bright cold day in April, and the clocks were striking thirteen”. In response, the machine penned: “I was in my car on my way to a new job in Seattle. I put the gas in, put the key in, and then I let it run. I just imagined what the day would be like. A hundred years from now. In 2045, I was a teacher in some school in a poor part of rural China. I started with Chinese history and history of science.” This is definitely an unorthodox response, but one that is certainly capable of providing a fresh perspective to the novel. Upon personal perusal of the key details that the machine introduces through this text, I could see that it talks about how the protagonist is envisioning its position in 2045, perhaps highlighting a key nature of AI and us as the makers; we are always looking into the future. In addition to this, the fact that the AI writes “a hundred years from now” is also quite important, showing how the protagonist is expecting to live to become a centurion, although this was quite the prospect at the actual time of writing of ‘1984’. I found this to be a great way of showing the versatility of GPT-2- how it can imbibe a brand new set of themes using a similar style of writing as the text it is given.
On the other hand, as with any writing source, there is always one key issue that has to be addressed- bias. Critics have deemed GPT-2’s writing style as “Trumpian”. This was shown when it was asked to write a follow-up to the line “Brexit has already cost the UK economy at least £80 billion since the EU referendum”, it began to go off on a tangent, drawing upon numerous warnings about UK’s economy and shedding most of the limelight on negative socio-economic impacts of Brexit. This is sure to beg the question as to whether this new AI can truly be a plausible source of information when it has such a severe sense of bias, and maybe this is something that needs to be worked on or improved before this technology can be used on a much larger scale. It is common knowledge that it is nearly impossible for a writer to be completely objective, but could GPT-2 prove to be the difference?
All in all, GPT-2 is a fantastic new invention, and despite its contentious style of writing, it has the potential to provide solutions to informatics-based issues and also introduce the world to a brand new style of writing. Authors could become antiquated in the future due to this technology, but if 2184 were to be released in the future by GPT-2, would it still be such a shame?
It would probably be fair to say that Stephen King didn’t get it all right with his aforementioned statement. Maybe he will change his quote after reading GPT-2, or just have the machine do it for him.
Sources:
Is writing on the wall for humans?