Compression could be machine learning's "killer app"

Pete Warden (previously) writes persuasively that machine learning companies could make a ton of money by turning to data-compression: for example, ML systems could convert your speech to text, then back into speech using a high-fidelity facsimile of your voice at the other end, saving enormous amounts of bandwidth in between.


Less exotically, ML is also used for "adaptive compression" algorithms that use ML-based judgments to decide how to compress different parts of a data-stream without compromising fidelity in ways that are perceptible by human observers.


Warden points out that companies already spend a lot of money on compression: vendors that want to sell ML-based compression systems would be asking for customers to switch who they spend an existing budget with, a much easier sell than convincing companies to spend money in an altogether new category.

One of the other reasons I think ML is such a good fit for compression is how many interesting results we've had recently with natural language. If you squint, you can see captioning as a way of radically compressing an image. One of the projects I've long wanted to create is a camera that runs captioning at one frame per second, and then writes each one out as a series of lines in a log file. That would create a very simplistic story of what the camera sees over time, I think of it as a narrative sensor.

The reason I think of this as compression is that you can then apply a generative neural network to each caption to recreate images. The images won't be literal matches to the inputs, but they should carry the same meaning. If you want results that are closer to the originals, you can also look at stylization, for example to create a line drawing of each scene. What these techniques have in common is that they identify parts of the input that are most important to us as people, and ignore the rest.


Will Compression Be Machine Learning's Killer App? [Pete Warden]


(via /.)

(Image: Cryteria, CC-BY)