Many of our day-to-day activities have been affected by artificial intelligence … so have you ever wondered what kind of knock-knock jokes or cookie recipes it would write?
Dr. Janelle Shane is an optics researcher with a Ph.D. in electrical engineering who spends her free time collecting the often humorous outputs from AI research projects. She maintains her menagerie of early AI blunders on her AI Weirdness Tumblr.
Her 2019 book “You look like a thing and I love you: How Artificial Intelligence Works and Why It’s Making the World a Weirder Place” is what we didn’t even know we were waiting for: a funny book about artificial intelligence. This work is a scenic railway tour of recent AI research presented with literary flair. She discusses — mostly — machine learning and neural networks.
You can buy the book on Amazon here. Your purchase helps support our ability to publish original speculative fiction.
This is definitely the new AI of deep-learning networks and generative adversarial networks, not the old AI of chess-playing machines and slow, deliberate robots. Two of the research projects cited are from 2019, and most of her examples are from no earlier than 2015. Janelle Shane does spend some time on non-neural-net types of machine learning like genetic algorithms and Markov processes, but mostly it’s all about neural networks. Ergo, AI is likely to see more of these funny blunders in the future.
Here’s a didactic — not funny — backgrounder on generative adversarial networks:
The title of Dr. Shane’s book, “You Look Like a Thing and I Love You,”- came from the output of a neural network trained with hundreds of examples of pick-up lines heard in singles bars. Among the many things she trained her neural net darlings to produce was innovative dessert recipes. A typical example is horseradish brownies. The author baked up a batch of horseradish brownies — in the eternal quest for scientific truth. She says they are even worse than what you would expect. Hard to believe.
This book is a celebration of the fresh new idea in artificial intelligence: it may well be impossible to build an intelligent machine, but it is possible to build machines that can learn to become intelligent. After all, that’s what nature did.
Edited by Amira Loutfi
After a long career in artificial intelligence research at IBM, JPL, NASA, and Hughes, Chuck Hand now lives on a five-acre farm in Massachusetts and puts in a solid three hours a day writing, reads voraciously, and cautions new writers not to use long sentences. He is also one of the founding members of MetaStellar. Contact him at [email protected].