Today's Paper State News Hutchinson 2024 LEARNS Guide Newsletters Opinion Sports Obits Games Archive Notices Core Values


by Philip Martin | May 23, 2023 at 4:42 a.m.

I am teaching myself to play bass.

Well, that's not quite true. I am learning how to construct rudimentary bass lines for the songs I write and record. I can do this with an electric six-string bass, but more often I use a regular guitar outfitted with a special pickup that converts the vibrations from the strings into a MIDI signal. (I can also use a MIDI keyboard if I want, but my keyboard skills are nil.)

It's not that complicated; you pop in some cables and fiddle with some settings in your digital audio workstation (DAW) and your guitar can sound sorta like a saxophone if that's what you want. I'm not really teaching myself to play bass; I'm not even trying to think like a real bass player. I just want to double the root note of the guitar chord an octave lower. It's a form of recreative play. I plunk along with the track and get a feeling about how far away I am from being a good bass player. Maybe I can get to a Sid Vicious or Stu Sutcliffe level of proficiency if I work really hard.

A couple of weeks ago, I got a piece of software that promised to listen to my tracks and create bass lines for me instantly. I was dubious--you have to tell the algorithm the key and the "genre" of the track, and I figured if it was really listening it ought to be able to figure that out on its own. But I gave it a try.

Boom. It's a better bass player than I am. The parts it comes up with are serviceable.

Best of all, you can have it switch things up, move notes around, change their velocity. You can add or delete notes. You can spend as much time as you want doctoring your two- or three-note bass figure. You don't need to be a virtuoso; if you can scoot a cursor around a screen and have some taste and patience, you too can piece together some complicated bass lines.

This is one way Artificial Intelligence has affected my life. It's been coming for some time--in 2016, Apple released an app called Music Memos that was able to analyze rhythm and chords of acoustic guitar and piano recordings and instantly add drums and a bass. It was surprisingly accurate, and I know some singer-songwriter types who leaned heavily on it. You could scratch out an idea in the middle of the night and later import it into Garageband or Logic.

Apple discontinued the app in 2021, probably because there aren't enough potential users to make it worth their while. Other companies stepped in and improved upon the product. The program I downloaded a couple of weeks ago is a much better bass player than the Music Memos algorithm. And now it's possible to make music simply by typing in a text prompt. The tech powers that be have created self-training plagiarism engines that can write B+ college essays and make Kendrick Lamar do a duet with Nina Simone at the drop of a prompt.

I'm not an alarmist. I'm cautiously optimistic about the ways AI is changing our lives. I think it's going to be great as a medical diagnosis tool.

And if you're a teacher, you better start making your students write in the classroom. On foolscap with No. 2 pencils. Because anything they bring in from home might have come from one of those plagiarism engines.

No doubt we will see deepfake mischief; impossible photos circulated as evidence and voices pirated to endorse ideas and causes their owners abhor. Already there are those who traffick in deceit, who sell the wishful rumors and lies repackaged as plausibilities. AI will be their ally.

AI is a tool that threshes facts; it churns through data and notices points of congruence. It processes massive amounts of data, looking for patterns to model in its decision-making. Most of the time, at least for now, puny humans supervise an AI's learning process, reinforcing good decisions and discouraging bad ones. Some AI systems learn without supervision--they will, for example, play a video game over and over until they eventually figure how to beat it.

And this is a problem. More information--more data points--does not always mean a better product. There are areas of human endeavor where it's not always best to know more facts. Tic-tac-toe is amusing until you realize it's always possible not to lose--to force a draw.

Baseball was a better game--that is to say it was a better entertainment product--before we discovered that the most efficient way to win games was to have pitchers expending maximum effort on every pitch rather than trying to throw a complete game. When we had hitters who were shamed by striking out and prioritized putting the ball in play, baseball was more exciting and fun to watch.

Basketball too was a better watch before analytics taught us the best way to play was to shoot three-pointers 40 percent of the time. The better way to play--from a competitors' point of view--is rarely the most aesthetically pleasing way to play.

The reason all the movies that end up as the best box office performers have numerals in their names is because the data proves these movies are more efficient vehicles for extracting disposable income from people looking for places to go to court affection than "Five Easy Pieces" or 'The Godfather" ever were. AI can help Hollywood make more lucrative movies, but not better movies.

But I think the humans win. Maybe it requires an act of faith to think that, but I just don't think we're all that interested in reading novels spit out by AI that's analyzed a hundred years of bestsellers or listening to synthetic approximations of what a Dolly Parton-Kendrick Lamar duet might sound like. Those things are interesting in an inside-jokey show-this-to-your-buddies sort of way, but I hope our species is still able to discern and appreciate the mystery at the heart of human-made art.

AI might be OK for churning out outlines for scripts for formulaic police procedurals or comic book movies, but absent the human element--the messiness of the private heart--things feel sterile and slick, cool to the touch. As we approach perfection, we reach an uncanny valley where everything is a little too neat, a little too on top of the beat.

I can transcribe a Jaco Pastorius bass solo in my new app, but when I play it back it lacks thunder and pain. It's just a series of pulses.

And if we fail to recognize and appreciate the difference, that's on us.

Philip Martin is a columnist and critic for the Arkansas Democrat-Gazette. Email him at

Print Headline: The limits of AI


Sponsor Content