Story telling: Computer vs Human
“Computers have been inserting themselves into activities long reserved for humans. They can play on a par with the best chest player in the world, the best poker players…but can a computer write a short story or a sonnet thats indistinguishable from what a human could produce?…Joe Palca is here to walk us through this. He first told us about a competition sponsored by computer scientists at Dartmouth Collect that tried to answer that question last year. What did the Dartmouth scientists want to get computer programs that could write sonnets or short stories? The idea was to see if they could understand human nature enough and science enough to pass something thats called the Touring test…If you can pass the Touring test it means you’ve written a programme that generates an output of sonnets that are indistinguishable from what a human sonnet writer could do…computer generated poems have rhyme, they have meter but they don’t mean anything. Narrative is a very difficult thing to program…You can get sensible phrases but then stringing them together to construct an overall narrative image is still quite a challenge….In the competition, i found there there were some by humans, that were such developed or insightful images, that the though that a computer could do that, i would be both amazed if It could, I would be terrified actually, because it would suggest a knowledge of human experience, and I don’t think machines could have or should have frankly…Story telling, I think, many people have said, is one of the things that is most human.”
— Joe Palca “Human or Machine: Can You Tell Who Wrote These Poems?” (2016)
The topic taken on for ‘Human Or Machine: Can You Tell Who Wrote These Poems?’ questions whether machines can produce written content identical to humans. The podcast attacks this question by assuming the various tasks computers and machines have the ability to accomplish today, and whether this means that they also have the capabilities to write “a short story that is indistinguishable from what a human could write.” This is discussed by Joe Palca and Professor Daniel Rockmore.
Palca argues that the point of the investigation into humans versus machines in the world of writing is to understand whether machines could understand basic human nature enough to produce content “indistinguishable” from humans. In the broader context of digital culture, this source serves particular relevance to the topic of generative texts in that it refers to the automated, generation of content by machines based on known inputs and generated by anticipated meaning. The known inputs in respect to this source is the wording computers have picked up on and learnt from humans as they create narratives. This is how machines generate written work. Here we notice this source to be relevant in the broader context of digital culture with its account for machine generative texts.
The source suggests that although machines have involved themselves into human activities for a long time, whether they are in fact able to produce written work that is identical to that of a human is not possible. This relates to the concept of Recurrent Neural Networks (RNNs) and how RNNs accept an input and provide an output. The output’s contents are influenced not only by the input you feed it, but also on the entire history of inputs you’ve fed it previously. However what this source argues is that there is only a certain extent to which computers are able to produce content, due to the lack of human knowledge and experience. So here, the RNNs for machines only exist because of human input. They do not produce unique content as we do.
An example from the source to prove that machines can not replicate human content is how there is an essential human element absent which enables us to distinguish what is computer generated content. This element being the “knowledge of human experience.” The artificial generation of content can be understood in digital culture by the deformance of generative content. For example when we write an essay on a text, we are deforming it when we review it in our own words. Retelling something in our own words is a form of deforming, and this is the basic structure machines work under when producing content such as sonnets as this podcast suggests. Arguably, deformance exists in the artificial creation of content because the producer has to know the value of what they are changing first before they can see the value of what you wish to change it to. There must be an understanding of the original text and the process by which it was deformed. We could argue that the original text in the instance of this source, is the knowledge of human experience, and it is fact that machines are to no extent human, nor have they had any life experiences, so with respect to digital culture, are instead assisting in the deformance of generative content with no actual first hand knowledge of what they narrate.
