Tech-wise at the moment we are living the same dichotomic situation we were experiencing two years ago. In March 2021, after Christie's auction of Beeple's purely digital work with a unique NFT set a record, selling at $69,346,250 USD, our collective attention focused on non fungible tokens. Soon NFTs became the reasons of abrupt creative joy, dubious marketing ploys and, more recently, the unlikely protagonists of some complex legal conundrums. Nowadays our attention is focused on Artificial Intelligence (AI) and on the benefits and risks that it may offer to humanity.
AI assisted text-to-image tools have unleashed thousands of new Instagram accounts posting bright and vivid images of fantasy worlds, garments, accessories and foods generated by an all-consuming AI. Artificial Intelligence-assisted writing tools such as ChatGPT turned instead into the main enemy of teachers and lecturers, afraid that students may cheat, something that resulted in some universities and schools banning it.
But, in the last few days, new benefits and anxieties of an AI-assisted world were reported. Australian federal Labor MP Julian Hill presented a speech today in Australia's parliament that was part-written by ChatGPT. Hill used ChatGPT prompts such as "please summarise recent media reports about students using artificial intelligence in Australia to cheat and explain why teachers are worried about this" and "explain in 2 minutes the risks and benefits to Australia from artificial general intelligence" to create parts of his speech (that has now become the first parliamentary speech part-written with the assistance of Artificial Intelligence).
Hill's speech partially focused on students using AI to write essays without understanding what they are writing and gaining an unfair advantage. The AI prompts Hill used to generate the text produced rather negative feedback, highlighting that the use of AI may cause job losses, perpetuate biases and discrimination and cause cyber-attacks and disinformation campaigns.
Other major dangers included further catastrophic risks especially if Artificial Intelligence surpasses human intelligence and in case its goals and motivations do not align with humanity's. In a nutshell, Artificial Intelligence may become too powerful, and may even transform warfare and have serious implications for national security. At the end of his speech, Hill urged to act together and urgently on Artificial General Intelligence (AGI), calling for regulations.
Now, while Hill's speech admittedly contained a highly dystopian content worth of a Netflix series, it is true that what scares many scientists is the fact that AI is doing very fast progresses: a superficial example of such progress is the precision of the results generated by text-to-image tools and applications. AI tools may still have some issues when it comes to creating human beings with five finger hands or coherent pictures, the images they produce are becoming more refined and visually intriguing.
Besides, if properly trained, Artificial Intelligence may lead to great progresses in different industries including healthcare and transportation.
There is actually a field in which Artificial Intelligence has proved useful – literature. A few days ago, Spain's National Library announced that Artificial Intelligence had uncovered a lost work from Spain's Golden Age. There are actually many plays from this age that, to this day, remain anonymous or are misattributed.
The discovery was done thanks to a project started six years ago by Germán Vega, a Golden Age literature expert at the University of Valladolid, and Álvaro Cuéllar, a researcher at the department of Romance studies at the University of Vienna.
Vega and Cuéllar launched Etso, a project that uses AI analysis to determine the authorship of Golden Age plays. Over 1,000 plays, most of them from Spain’s National Library, were digitally transcribed using a platform, Transkribus, trained to identify and understand 3m words. Then a programme called Stylo compared their language and style with the 2,800 digitised works by 350 authors in the Etso database.
While analyzing a text entitled "La francesa Laura" (The Frenchwoman Laura), a minor play about love and jealousy catalogued as an anonymous work, Etso noticed similarities between it and another 100 works by Spanish classical dramatist, poet and priest Lope de Vega.
Researchers then analysed the results and a further analysis of plots, character names, metre and style, confirmed the analysis. The discovery allowed therefore researchers to confirm the attribution and even date it for contents and style between 1628 and 1630.
Four years ago, Etso and Stylo were already used to established that 17th-century play "The Nun Lieutenant" (about Catalina de Erauso, who escaped a convent to become a cross-dressing soldier in the Americas) was written by a Mexican dramatist Juan Ruiz de Alarcón. While these cases refer to minor discoveries, these can still be considered as important landmarks.
Attributing a literary work to an author is a long process: researchers have to read hundreds, if not thousands, of texts, take notes, consult works in libraries based in different countries and, at times, compare texts with old manuscripts. A well-trained AI could therefore prove useful and reduce research times. Obviously this is not yet a straightforward process and it must be tuned in, but it will definitely get better with time.
Which takes us to the consideration that, just as Artificial Intelligence can prove useful to highlight the attribution of a literary text, it may be helpful in attributing works of art, but also in shedding light on some fashion archives. Artificial Intelligence-led projects in fashion and textile museums and institutions may indeed help us attributing creations to a designer or allow us to come up with unusual comparisons.
This obviously means that we will have to train Artificial Intelligence with well-documented materials and images and this is not always possible as the Internet is full of mistakes and even some databases are riddled with errors (it is not rare to find pictures mistakenly attributed: Wanda Wulz's "Io più gatto" (Cat + I) is often incorrectly attributed to Man Ray, especially on Pinterest boards; photographs of Anita Ekberg in "Boccaccio '70" are often listed online and even on certain school books as taken from Fellini's "La dolce vita" - Hill may be scared that AI will surpass human intelligence, but in some cases this has already happened, but that's probably because we have often stopped using our intelligence to ask ourselves questions and we do not fact check the notions we find online ...).
But training Artificial Intelligence with more precise databases - fashion and textile ones as well - is not impossible and it will happen as time passes. So let's not doubt we may be able to rein in the power of AI and use it for our researches, and let's consider AI as a support tool, rather than a scary entity that may turn against us à la Megan, the artificially intelligent doll who develops self-awareness in Gerard Johnstone's film "M3GAN" (2022).
Comments