Swearing Computers – AI’s Challenge of Toxic Degeneration

Neural toxic degeneation at AI2 and UW

Allen Institute for AI and University of Washington researchers, top row from left, Samuel Gehman, Suchin Gururangan, Maarten Sap, and bottom row from left, Yejin Choi, Noah A. Smith. (AI2 Photo)

After defeating the all-time human champions of the game show, Jeopardy in 2011, IBM’s Watson supercomputer was scheduled for an upgrade. Wanting to give it access to a broader vocabulary, its programmers introduced the AI to the Urban Dictionary, a web-based collection of cultural phrases and slang. Unfortunately, because of this, Watson was soon swearing profusely.

Something similar happened in 2016, when Microsoft introduced it’s new Twitter chatbot, Tay, to the world. Based on the persona of a teenage girl, the program devolved into a racist neo-Nazi after less than 24 hours on the internet and had to be removed.

Why do natural language AIs frequently degenerate in this way, cursing and spouting biased, racist and otherwise toxic messages. I recently explored these issues for Geekwire, speaking with a team of researchers from the Allen Institute of Artificial Intelligence and the University of Washington who are exploring this problem. While the problem is very challenging, their work may help address a problem that is becoming increasingly important as we increasingly rely on AIs to communicate with us using natural language.

A New Definition of Intelligence

Intelligence: An emergent system’s ability to respond to its environment in order to improve its conditions, perpetuate itself and maximize its future freedom of action. – Future Minds: The Rise of Intelligence from the Big Bang to the End of the Universe by Richard Yonck

In writing my latest book, Future Minds: The Rise of Intelligence from the Big Bang to the End of the Universe, it was essential to properly define intelligence in an appropriate context. However, my research led me to over 200 different definitions of this nebulous word, most of which were overly restrictive or naively broad. This is due in large part to intelligence being what AI pioneer Marvin Minsky called a “suitcase word”, a term he used to describe words having many meanings and associations. In a 1998 interview with Edge, Minsky said of suitcase-words (like ‘intelligence’, ‘intuition’ or ‘consciousness’): “all of us use these to encapsulate our jumbled ideas about our minds. We use those words as suitcases in which to contain all sorts of mysteries that we can’t yet explain.”

Because of this and for reasons I expand on throughout my book, I finally arrived at a broad definition that encompasses the ideas I’d been exploring.

Intelligence: An emergent system’s ability to respond to its environment in order to improve its conditions, perpetuate itself and maximize its future freedom of action.”

This allows intelligence to be seen much more as an almost inherent property of the universe, something that isn’t limited to a single species or substrate, but which is an ongoing optimization in those systems that are able to successfully perpetuate themselves into the future. Based on this, while it may take considerable time, it seems likely that advanced technological intelligences will one day become a reality.

Parsons School of Design Guest Lecture

I recently had the pleasure of giving a guest lecture at Parsons School of Design. My presentation, “Designing for a Differently Intelligent Future” is drawn from parts of my latest book, Future Minds. It was a lot of fun and as always, I think I enjoyed the Q&A the most. A big thank you to friend and trend analyst Tim Stock for inviting me to speak with his class. Let’s do it again!

Future Minds Book Launch

It was almost exactly half a year ago that the official book launch for Future Minds at Town Hall Seattle was postponed. We expected to have a packed house, but I cancelled it two weeks earlier as it became apparent that any sort of large gathering would be irresponsible given the impending pandemic. Six months later, it was finally time to return to THS, albeit virtually, for the “official” launch party! A huge thank you to Wier Harman, Shane Unger and everyone at Town Hall Seattle, as well as the evening’s official bookseller, Third Place Books.

Amazon Alexa Prize Finals – Team Emora Wins

Congratulations Team Emora! Amazon announced today that Emory University and their Emora socialbot are the winners of the 2020 Alexa Prize. Second Place was awarded to Stanford’s Chirpy Cardinal bot and Third Place was Alquist developed by the team from Czech Technical University.

Emory won with a 3.81 average rating (out of a possible 5.0), earning them the $500,000 top prize. Each of the past two years’ winners have exceeded the average rating of the previous year’s winner. The eventual goal of the competition is to achieve a 4.0 rating while conversing with a human interactor for a minimum of twenty minutes.

It’s been an honor to be a part of this innovative competition – my third year as an Amazon #AlexaPrize finalist judge. I look forward to seeing what next year’s competition brings and what all of the competitors do with their amazing talent in the future!

Future Minds – Emerging Intelligence Interview

Not surprisingly, the book launch of Future Minds at Town Hall Seattle had to be postponed, a decision I wholeheartedly support in the interest of public safety. But in lieu of that presentation (at least for now), the good folks at THS did an interview with me to offer a taste of what my book and presentation are about. You can read my interview in their blog, the Town Crier. Find out why our world is rapidly becoming more and differently intelligent and what this could mean for the future!