Explore a blog that specializes in synthesizing poems composed by AI thơ

There are countless blogs online dedicated to posting poetry, but Curated AI is more special: it’s a library of computer-generated poems. If the Turing test is an objective way (although this is still controversial) to measure the “humanity” of a computer, then poetry is an “objective” way of cheating.

“When the computer has learned enough, writing poetry is just a small thing,” according to Karmel Allison, author of Curated AI.

Allison said CuratedAI is the result of a side project of hers on the topic of the poetic ability of artificial neural networks. As someone who has written poetry for many years, she is struck by the way computers create coherent, even beautiful lines of poetry. This joke-sounding project turned out to be very real to her.

“I think the meaning of a work lies in the reader, not in the writer. You can explain what data the computer learns on, how it works – but you can’t get the idea of ​​it – or you can call it the idea of ​​an algorithm, but read the work directly to understand out is better.“

Allison has been a programmer and data researcher for a decade. She is also working with biological data at a biotech company with the aim of developing a program to sequence the structure of DNA. When asked which is easier to write: the source code of the algorithm for poetry or poetry, she took a moment to answer: “I think writing good poetry is difficult, right? I have never written a poem myself, nor have I had a computer write a poem that can be called good. But thanks to the computer, you can write more poetry. Just have to be more selective.“

Sometimes writers rambling and editors have to cut those spots. The same goes for poetry, written by humans or machines. But machines never fail to come up with words: Allison’s Deep Gimle I algorithm can access a dictionary of 190,000 words to create verses in less than a minute. The publisher McGraw Hill estimates that the average person uses a vocabulary of about 10,000 words when writing, with Shakespeare plays alone having a vocabulary of about 33,000

The number 190,000 is not something special. Allison concedes that it is possible that the algorithm has some unexpected lexical choices, even though it is programmed to avoid doing this. The reason for this is also the reason why not many people understand most of that vocabulary.

Computer-generated works of art are responded to with nods from real artists, an amazing achievement for many reasons. Creativity no longer has the same meaning as before. Instead, creations are now marvels from a computing perspective.

Try reading the poem “Crazy” from the author of “Gimble” and reflect on the gap between the ability of humans and machines to synthesize an impressive work on an emotional and very “human” subject. :

Original:

Translate:

If you want to put your AI to the test, CuratedAI is accepting cards!

According to PopSci