Facebook’s mea culpa, party parrots, and is AI art real art? (Issue 11)

Welcome to Issue 11 of the All Turtles newsletter, sent on January 18, 2018. Each week, we bring you carefully chosen news and analysis about AI, startups, and happenings at the All Turtles startup studio. If you like this newsletter (we hope you do!), please subscribe or share with a friend.

Hello again! This week, our newsletter picks inspired us to think about the future of art and human creativity. We also considered the simultaneously wonderful and nefarious ways AI is being used by social media. And in our podcast, we learned how a company whose workers are fully-remote can create a terrific company culture. We hope you enjoy this issue of our newsletter. If you know someone who might enjoy it, please forward this newsletter to them.


Is art created by a machine still art? Even if an AI creates a “masterpiece,” is it valuable when a thousand more are just a click away? Asking a computer to make something art-like is easier than ever, but it remains to be seen what that will do to our appreciation of creativity (and of the creators themselves). Computer-aided art is giving rise to new thought experiments — such as whether we’d pay the same premium for a Spotify subscription when the music we listen to is machine-generated — which are challenging long-held beliefs about how we value the arts.

Read: Is Art Created by AI Really Art? (Scientific American)

Party parrots

We check in with Octane AI cofounder and COO Ben Parr about bots and ecommerce, and the surprising ways parrots may help build company culture. Along the way hosts Phil Libin, Jessica Collier, and Blaise Zerega question the importance of boards of directors in a segment called, “Maybe it’s kinda bullshit?” They also discuss whether design can address ethical concerns for AI and what the demise of Facebook M means for chatbots.

Listen: All Turtles Podcast Episode 11: Batteries Not Included (iTunes)

Move slow, fix things

It’s not new and it may not top AI’s 2018 sexy list, but Facebook’s newsfeed is the AI product that many people around the world interact with most. This also means that their newly announced plan to overhaul the newsfeed may be the biggest dial-back of an AI product, well, ever.

The newsfeed has evolved to be a one-stop media shop of news and goofy videos shared by friends. But this announcement is a significant mea culpa from Facebook, which has been under fire for the spread of fake news and the ease with which media influence can be bought. The company plans to show more posts from friends and fewer posts from businesses, brands, and media, showing that even the most well-intended algorithms can backfire.

Read: Facebook’s startling new ambition is to shrink (The Verge)

/giphy drowning

While Facebook is working to alleviate product pain, Slack is trying to use AI to avoid it altogether. Slack fatigue is a real phenomenon, and they know it. Their challenge: in a sea of messages darting around a company, how can the service smartly highlight the ones that are most likely to be important to you? Here’s a look at their interesting approach to a challenge that may have a real impact on your own productivity.

Read: Slack Hopes Its AI Will Keep You from Hating Slack (MIT Technology Review)

Be careful what you wish for…

Continuing on the theme, Facebook’s upcoming changes are a de facto admission of how big a problem fake news has become. The discussion of current events and social issues seem almost permanently distorted by fictitious journalism and its dissemination by botnets and bad actors. Somewhere in this problem space lies a very good product that uses natural language processing to combat fake news. One programmer’s attempt at it yields a 95-percent success rate at detecting “non-real” articles, but getting there wasn’t pretty. Here’s his story.

Read: I trained fake news detection AI with >95% accuracy, and almost went crazy (Medium)

We’d love to hear from you. Simply email us with feedback or to suggest articles, videos, and papers for future newsletters. Sending an email also works for questions about practical AI, company formation, or product development that you’d like answered in a future podcast episode. We devote a portion of each show to listener questions and comments.

Thanks for reading.