How to Floss Your Youtube Algo

I once had a friend named Rick Barraza. He works at Microsoft and the last time I talked to him, about three years ago in Redmond, he was deleting all of his social media accounts. One of his chief concerns as a member of what was then the Responsibility in AI group (?) was with the use of AI by media content distributors to recommend new content based on your previous content preferences. A side-effect / primary goal (depending on your hermeneutical suspicion level) of these “recommendation engines” is to actually steer you toward more and more extreme content which increases your viewing time with the content you are being fed. Recommendation engines can actually modify your outlook based on the content it feeds you. This effect has been identified as a root cause of the increase in both left wing and right wing extremist outlooks over the past half decade.

Rick is now offline and lost to me but that last conversation continues to echo through the years. When I pull up Youtube I often find some  odd things showing up in my recommendation list. At a minimum, I don’t want these things to show up when my children come into the living room. So I came up with a playlist of random and neutral content that I can run through my Youtube account to normalize it and get rid of these outliers.

Then a week ago it occurred to me that I could use these regular algo flossings to also do a bit of mental flossing. To the extent that Youtube’s algorithms remotely alter my emotional rhythms and outlook and tastes, it occurred to me that I could piggyback on Youtube’s hack of my mind in order to hack myself. By carefully modifying the model that Youtube has created of me, I can perhaps perform the highest form of self-care.

I wanted to give you a taste of some of the pieces I am using to shift the model in case it helps you. I just run the full list through Youtube every few days and walk away. I’ve noticed that the content recommendations I now get out of Youtube are much more edifying. At a minimum, it is much less embarrassing for me, personally, when my children see it.

Adam Neely’s music theory video about The Girl From Ipenema is extraordinary and profound. But that isn’t why it works well for flossing. The exploration of the gentrification of Brazilian music in a way that created cross pollination with American bebop and an unexpectedly deep and ambiguous musical chord structure creates a pleasant nexus of 50’s themes, mathematics, interesting graphics and lots of variations on the same song. This creates a latticework of connections to other affirming content that work together to drive out the less pleasant things that may inadvertently show up on your recommendation list.

Tony Zhou’s Every Frame a Painting is a series about film technique and appreciation. Tony is a video editor who came to this from creating content for The Criterion Channel (which by the way is a magnificent first year course on the building blocks of filmmaking if you watch them all). The long shot video, in particular, shifts your model toward other videos about the one shot, which is almost always received as a virtuoso film technique that is used in both artsy and popular movies. Playing it a few times in your playlist will surface still shots from movies that will tend to make your YouTube recommends much more visually appealing.

Emmanuel Levinas is one of the greatest unknown modern philosophers. He studied in the French / German existential tradition but is most famous for his ethics. He famously said that Ethics is first philosophy, which is a complete re-orientation from Heidegger’s ontology as first philosophy, Descartes’ epistemology as first philosophy or Aristotle’s metaphysics as first philosophy. His thinking about The Other has made its way into contemporary culture, although his understanding of The Other is much different from the way popular culture has received it. I used to use Slavoj Zizek, the neo-Trotskyite, as part of my flossing regimen but have found that ironic humor can be problematic when attempting to perform algorithmic hygiene and can introduce some mutations of the recommendation model that are not optimal. More on this in a future post.

Three minutes of Ben Hogan’s golf swing is a bit of a palate cleanser. It is a perfect combination of excellence and boredom, which is what we are after when creating a flossing playlist.

Anuja Kamat’s What is a Raag? besides being lovely and informative fixes the implicit tendency of algos to whitewash. Ben Hogan’s back swing is like white bread to sop up left over gravy. It sets up Anuja Kamat’s follow through which attracts links to high quality content outside of this safe zone. Just as having too much irony can be a problem in our playlist, and invites erratic content, a narrowness of horizons can similarly distort the recommendation model in undesirable ways.

Preparing banh xeo (Vietnamese crepe) is a subgenre cross between cooking instruction and ASMR. While almost any banh xeo video will probably work, this one is especially interesting for the amount of time spent on establishing credibility and authenticity.

At first I was worried that Khruangbin’s performance of Maria Tambien, August 10 and White Gloves for NPR’s Tiny Desk Concert series would open my algo up to radical leftist politics, which while not necessarily bad in itself can lead to a self-perpetuating extremist slope in the model, but instead the moderating calmness of public broadcasting is what ultimately influences the algorithm the most. This neo-liberal tendency is actually perfect for algo and mental flossing.

Other content which would seem similar on the surface, such as TED talks and TEDx talks, anecdotally have those sliding side-effects toward cesspool media and should be avoided at all cost when cleaning up your video profile. If you must watch a TED talk, try to watch it in your browser’s “private” mode (though this isn’t guaranteed to always remove the deleterious influence on your recommendation model, so be careful).

The last item I want to share with you from my Youtube hygiene playlist is a talk about 5 dangerous ideas entitled Crafting Delight by Rick Barraza. The talk itself is nothing special and you should probably just leave it playing with the sound off while you go do something else. I believe (though I cannot prove it), however, that embedded in the video are subharmonic or subliminal signals recognizable to the Youtube algorithms that effectively reset them so they are less predatory and potentially even beneficial.

Again, I can make no empirical claims about whether the Crafting Delight video is actually an encoded vaccine for predatory recommendation engines. All I know is that it seems to work for me. I also cannot say whether there are or are not other algo vaccination videos on Youtube pretending to be recorded night club performances, travelogs, or even software programming instructional videos. I do get tips from time to time though about the beneficial mental and algorithmic side-effects of various media. What I suggest is that you try these out and if they work for you, then hit me up for the rest of the playlist.

If these videos have helped you to de-program, please let me know in the comments so I can weight them correctly for my future counter-algo algo training data sets.

The Great AI Awakening

bathers_and_whale

This is a crazy long but nicely comprehensive article by the New York Times on the current state of AI: The Great AI Awakening.

While lately I’ve been buried in 3D interfaces, I’m always faintly aware of the way 1D interfaces (Cortana Skills, Speech as a service, etc.) is another fruit of our recent machine learning breakthroughs (or more accurately refocus) and of how the future success of holographic displays ultimately involves making it work with our 1D interfaces to create personal assistants. This article helps connect the dots between these, at first, apparently different technologies.

It also nicely complements Memo Atken’s Medium posts on Deep Learning and Art, which Microsoft resident genius Rick Barraza pointed me to a while back:

Part 1: The Dawn of Deep Learning

Part 2: Algorithmic Decision Making, Machine Bias, Creativity and Diversity

There’s also a nice throw away reference in the Times article about the relationship between VR and Machine Learning which is a little less obscure if you already know Baudrillard’s Simulacra and Simulation which in turn depends on Jorge Luis Borges’s very short story On Exactitude In Science.

If you really haven’t the time though, which I suspect may be the case, here are some quick excerpts starting with Google’s AI efforts:

Google’s decision to reorganize itself around A.I. was the first major manifestation of what has become an industrywide machine-learning delirium. Over the past four years, six companies in particular — Google, Facebook, Apple, Amazon, Microsoft and the Chinese firm Baidu — have touched off an arms race for A.I. talent, particularly within universities. Corporate promises of resources and freedom have thinned out top academic departments. It has become widely known in Silicon Valley that Mark Zuckerberg, chief executive of Facebook, personally oversees, with phone calls and video-chat blandishments, his company’s overtures to the most desirable graduate students. Starting salaries of seven figures are not unheard-of. Attendance at the field’s most important academic conference has nearly quadrupled. What is at stake is not just one more piecemeal innovation but control over what very well could represent an entirely new computational platform: pervasive, ambient artificial intelligence.

 

When he has an opportunity to make careful distinctions, Pichai differentiates between the current applications of A.I. and the ultimate goal of “artificial general intelligence.” Artificial general intelligence will not involve dutiful adherence to explicit instructions, but instead will demonstrate a facility with the implicit, the interpretive. It will be a general tool, designed for general purposes in a general context. Pichai believes his company’s future depends on something like this. Imagine if you could tell Google Maps, “I’d like to go to the airport, but I need to stop off on the way to buy a present for my nephew.” A more generally intelligent version of that service — a ubiquitous assistant, of the sort that Scarlett Johansson memorably disembodied three years ago in the Spike Jonze film “Her”— would know all sorts of things that, say, a close friend or an earnest intern might know: your nephew’s age, and how much you ordinarily like to spend on gifts for children, and where to find an open store. But a truly intelligent Maps could also conceivably know all sorts of things a close friend wouldn’t, like what has only recently come into fashion among preschoolers in your nephew’s school — or more important, what its users actually want. If an intelligent machine were able to discern some intricate if murky regularity in data about what we have done in the past, it might be able to extrapolate about our subsequent desires, even if we don’t entirely know them ourselves.

 

The new wave of A.I.-enhanced assistants — Apple’s Siri, Facebook’s M, Amazon’s Echo — are all creatures of machine learning, built with similar intentions. The corporate dreams for machine learning, however, aren’t exhausted by the goal of consumer clairvoyance. A medical-imaging subsidiary of Samsung announced this year that its new ultrasound devices could detect breast cancer. Management consultants are falling all over themselves to prep executives for the widening industrial applications of computers that program themselves. DeepMind, a 2014 Google acquisition, defeated the reigning human grandmaster of the ancient board game Go, despite predictions that such an achievement would take another 10 years.