Will ASMR Survive Machine Learning?

Original article was published on Artificial Intelligence on Medium

Much of the ASMR world extends beyond interpersonal closeness, into a deeper, more erotic intimacy. Which brings me back to the ethereal nymphs nibbling on my ears. Many of the most viewed ASMRtists, like FrivolousFox, Pelagea ASMR, and ASMR Cherry Crush, have converged on styles that burgeon with ‘mouth sounds’, including kissing, nibbling, and sucking on their binaural microphones. Accompanying visuals include close-up lips, doe eyes, and gentle stroking hand movements.

ASMR Cherry Crush attends to your ears. Video from Youtube.

This sub-genre undeniably draws some of its power from its erotic potential, amplified by a languid, unhurried delivery. That erotic potential results in unwelcome attention of the kind the Internet seems so reliable at delivering. Cherry Crush emphasizes on her YouTube channel that her ASMR “is NOT for fetishization of sexualization purposes”. Yet it doesn’t take an overactive imagination to see why such a disclaimer might become worth posting.

There is so much more to ASMR, and ASMR intimacy, than this particular form of eroticism. Ephemeral Rift’s dark experimental triggers and Goodnight Moon’s quirky storytelling depart from the heavily gendered whispering intimacy that seems to dominate the art form right now.

Ephemeral Rift’s Professor Clemmons will trigger the tingles while he treats you like a plant-human hybrid.

Aural porn

The success of obliquely sensual mouth sounding and ear-nibbling ASMRtists has caught the attention of pornographers. Adept at monetizing almost any erotic outlet, they are throwing their resources and best talent at ASMR. They repurpose explicit videos, remixing the soundtracks, to deliver both the ASMR tingles and the other kind. Angela White filmed new sex scenes for the genre. A naked Rockey Emerson sits topless in another video, eating a packet of macarons.

Despite the protests of many rusted-on ASMR fans, regular and X-rated ASMR seem almost indistinguishable to many listeners. Soundtracks play an enormous part in pornographic films, from the sounds of bodies coming together to the exaggerated gasps, sighs, and “Oh my God!” expressions of the performers. New artists like Brasileira Maru Karv emerge daily to fill the gap, starring in explicit solo videos that prioritize ASMR triggers over the sexual acts themselves.

Fertile ground for machine learning

While ASMRtists start to feel the sharp elbows of Big Porn, a bigger challenge to their artform might come from another direction. Artificial Intelligence, and especially machine learning, seems perfectly suited to exploring and creating ASMR. In so doing, it could make tingle-generation a largely computer-driven pursuit.

Chances are that ASMRtists and hobbyists have only mined the surface deposits of ASMR triggers

In recent years, music made with the help of artificial intelligence leaped from futuristic prediction to audible reality. Projects like Google’s Magenta and OpenAI’s MuseNet lead the public charge toward AI-based musical creation. For now, ML algorithms learn from the vast body of music already created by human composers and musicians. They discover patterns in the music and learn to predict chords and notes that might follow others, sometimes stumbling into never-before-heard melodies.

ASMRtists and hobbyists have probably only mined the surface deposits of ASMR triggers. Their finds, recorded and published on the Internet, suit machine learning. The structures of ASMR tracks are simpler than music, and listener responses lend themselves easily to measurement. At the Tingle Science website, users can log their responses to ASMR videos by holding down a button while the feeling lasts. It should not prove difficult to track heart rate or skin conductance via existing smartwatches, generating mountains of data. Any such measures provide excellent data for machine learning.

There exists no reason for AI-generated ASMR triggers to remain constrained to the world of natural sounds. From data on existing sounds, algorithms should be able to predict — and try out — new sounds that elicit ASMR even more effectively. From those sounds, and people’s responses to those sounds, machine learning ASMR could walk step-by-step into the terra incognita of sounds never before heard by human ears, dramatically expanding ASMR repertoires.

That will likely lead to super-normal stimuli, entirely new sounds that evoke responses stronger and more reliable than present-day ASMR. Endowed with such capacities to hack into our intimacy pathways, platforms that get the Machine Learning of ASMR right will almost certainly exert a far more effective hold on human attention spans than our existing technologies do today, keeping us listening and engaging by feeding us the tingles.

ASMR might be even more ripe for digital disruption than music. Photo by Spencer Imbrock on Unsplash