Platypus Header

Platypus Innovation Blog

15 February 2019

GPT-2: A new AI with amazing creative writing skills

This language AI is really amazing. Let's cut to the unicorns:

The AI was given a creative writing task: Given a prompt, and asked to continue writing the story. The example reported has been cherry-picked (i.e. the researchers filtered out the duds and are showing off the best output), but is still very very impressive.

Prompt:
In a shocking finding, scientist discovered a herd of unicorns living in a remote, previously unexplored valley, in the Andes Mountains. Even more surprising to the researchers was the fact that the unicorns spoke perfect English.
Model completion (machine written!)
The scientist named the population, after their distinctive horn, Ovid’s Unicorn. These four-horned, silver-white unicorns were previously unknown to science. 
Now, after almost two centuries, the mystery of what sparked this odd phenomenon is finally solved. 
Dr. Jorge Pérez, an evolutionary biologist from the University of La Paz, and several companions, were exploring the Andes Mountains when they found a small valley, with no other animals or humans. Pérez noticed that the valley had what appeared to be a natural fountain, surrounded by two peaks of rock and silver snow.
Pérez and the others then ventured further into the valley. “By the time we reached the top of one peak, the water looked blue, with some crystals on top,” said Pérez. 
Pérez and his friends were astonished to see the unicorn herd. These creatures could be seen from the air without having to move too much to see them – they were so close they could touch their horns. 
While examining these bizarre creatures the scientists discovered that the creatures also spoke some fairly regular English. Pérez stated, “We can see, for example, that they have a common ‘language,’ something like a dialect or dialectic.” 
Dr. Pérez believes that the unicorns may have originated in Argentina, where the animals were believed to be descendants of a lost race of people who lived there before the arrival of humans in those parts of South America. 
While their origins are still unclear, some believe that perhaps the creatures were created when a human and a unicorn met each other in a time before human civilization. According to Pérez, “In South America, such incidents seem to be quite common.” 
However, Pérez also pointed out that it is likely that the only way of knowing for sure if unicorns are indeed the descendants of a lost alien race is through DNA. “But they seem to be able to communicate in English quite well, which I believe is a sign of evolution, or at least a change in social organization,” said the scientist.
Wow.

And: the AI was entirely self-taught. It has a structure designed for sequence learning, which is then trained on the text from millions of web pages. It learns to predict the next word when reading. The impressive knowledge of language patterns and story structures shown above is all learned from the data.

As the researchers note, this level of AI has a lot of applications - good and bad. So they are not releasing the full model yet, asking the AI and wider society to consider how we manage this technology.

I read this yesterday. Still processing it with my jaw on the floor.

Naming things is an important part of humanising them, so the researchers have called this system GPT-2. See https://blog.openai.com/better-language-models/ for a summary of GPT-2 and a link to the technical paper. The neural net architecture is not given, but the paper and partial code suggest it may be surprisingly simple and generic, though large and expensive to train. Spoiler alert: it's not an LSTM - long short-term memory, the neural net architecture which has ruled NLP work for the last few years. It uses an attention-based short term memory in an encode-decode setup called a Transformer. Though attention functions do have some common ground with the memory-gates of an LSTM. So it's evolution not revolution. Except there's a point where evolution becomes revolutionary.

By Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever from OpenAI.com

Good-Loop Unit