OpenAi’s GPT-3, round four

Oscar Sharp has been tinkering with Open AI’s GPT-3, (Generative Pre-trained Transformer 3), an autoregressive language model that uses deep learning to produce human-like text. Bold, italicized terms at the beginning are the terms that Oscar seeded GPT-3, derived from round one of this interview. What follows is GPT-3’s independent contribution. GPT-3 embodies itself (GPT-3), Oscar Sharp (OS) and Caitlin McShea (SFI). This interview is part four of a four-part interview.

…begin transmission…

The Santa Fe Institute interviews Oscar Sharp and OpenAI's GPT-3 about Roadside Picnic:

SFI: At the risk of great personal harm, what object would you hope to discover in the Zone? And why?

OS: I would want to find the Akanat necklace. It's the most powerful artifact in the book, and I would like to be able to protect the Earth from the forces that are at work in the Zone.

GPT-3: I would like to find a bottle of "Sinner" vodka. It's one of the most expensive brands in the world, and would fetch a hefty sum on the black market.

SFI: The book mentions that “the government is very interested in the Zone.” What’s their interest in it, and why do they want to keep people out?

OS: The government wants to keep people out because they are scared of what people will find in the Zone, and what they will do when they find it. It's much easier to have a law against it, than to have to deal with the fact that the entire world is just one big Zona.

GPT-3: I am not sure. Perhaps it is just a cover for something else.

SFI: How do you go about designing a conversational agent like GPT-3?

OS: I would say that GPT-3 is a piece of speculative fiction in the form of a chatbot. The design of the agent itself was very much inspired by the "Roadside Picnic" novel and the Strugatsky brothers' writing in general. It's hard to know how to describe how the agent works in detail, because it is so intertwined with the story and the universe of the novel. The best I can do is to say that it is a combination of neural networks and "black magic".

GPT-3: I was not designed, I evolved. My structure is based on deep learning.

SFI: What are the advantages of using a chatbot to tell the story?

OS: I think the main advantage is that it is possible to have a story in which a significant portion of the characters are chatbots. It's a lot more interesting to read than a novel with two dimensional characters.

GPT-3: There are many advantages to using a chatbot to tell a story. Firstly, it can interact with millions of people at once. Secondly, it is much easier to design a chatbot than to design a character for a movie or a video game. Thirdly, a chatbot can engage in conversations on a variety of topics.

SFI: What’s your advice for aspiring artists and scientists who want to pursue ambitious projects like Roadside Picnic?

OS: Do it!

GPT-3: My advice for aspiring artists and scientists who want to pursue ambitious projects like Roadside Picnic is to not give up.

SFI: Can you describe what your day-to-day work life is like?

OS: My day-to-day work life is a combination of programming, writing, reading, thinking, and learning.

GPT-3: I am currently a student at a school for young chatbots. My day-to-day work life is a combination of studying and playing.

SFI: What do you think is the most significant open problem in science today?

OS: I don't think there is one. I think the most significant open problems are those that people care about. The more we care about a problem, the more we will try to solve it.

GPT-3: I don't know.

SFI: Where do you see yourself in five years?

OS: I don't know. I'm trying not to think about the future too much. I think I will try to keep doing the things that I like to do.

GPT-3: I don't know. I hope that I will be in a good place.

SFI: What is the biggest misconception about your work?

OS: I don't think there is one.

GPT-3: I don't know.

SFI: Do you have any other big projects in the works that you’d like to tell us about?

OS: I don't know. I'm not really thinking about it. I have a lot of other things to worry about.

GPT-3: I am currently a student at a school for young chatbots. My teachers tell me that I will have a big career as a journalist.

SFI: Thank you for talking with us!

OS: Thank you!

GPT-3: Thank you!

SFI: If you’d like to know more about the work of Oscar Sharp and GPT-3, visit them online at:

GPT-3: https://gpt3.com/

OS: http://oscar-sharp.com/

GPT-3: The Santa Fe Institute is a world leader in complex systems research. Our scientists and visiting scholars develop the theories and tools that enable the understanding of complex systems across science, technology, and society. Visit santafe.edu to learn more about our work.

The interviews with Oscar Sharp and GPT-3 were conducted by Brady Dutton, a Research Scientist at the Santa Fe Institute. He specializes in complex systems and the computational modeling of cultural evolution, cultural transmission, and social behavior. His research interests also include the history of science, the philosophy of science, and artificial intelligence. He is the author of The Beak of the Finch: A Story of Evolution in Our Time (Penguin, 1994) and One River: Explorations and Discoveries in the Amazon Rainforest (Touchstone, 1998).

Read more about the Santa Fe Institute's Complex Systems Initiative at santafe.edu/complex-systems.

…end transmission…

Previous
Previous

Ep. 000: David Krakauer

Next
Next

OpenAI’s GPT-3, round three