can AI Teach us about music?

By Jess Santacroce
Music Writer, Phoenix Radio

Any music person will tell you that no form of AI will ever replace a human being writing music in response to something they experienced or learned. AI is also a poor substitute for a human writer to create a musician’s publicity and promotional materials, as AI generated writing is often bland and ordinary, and no bot is ever going to care whether or not what it wrote actually helped your band. The same holds true for replacing the person who designs your album cover with AI. You are not going to get the same quality, and you will get absolutely no care. But what about music teachers and promoters? Can AI teach us music skills, teach music history and current events, keep us up on the local music scene,  or promote music appreciation? 

Experiment #1: Guitar lessons 

I have been a professional artist for more than twenty years, but I am a novelist, content writer, and writing teacher, not a professional musician. At age fifty, I have sung as a hobby for more than forty years, but I have a good voice for a hobbyist, not a professional level voice in terms of quality. Decades ago, I briefly played guitar, but while I don’t think I was my teacher’s worst student, I was strongly encouraged to be a singer….not so much a guitar player. 

The above paragraph was entered into chat gpt with the question, “Can you teach me to play guitar?”

Chat gpt responded that it would be delighted to teach me guitar. It told me I already have a few major advantages, calling me a “seasoned artist with creative discipline,” and stating “You have the artistic patience and sensitivity to approach practice meaningfully, not mechanically.” 

It followed that up with a six month lesson plan broken down into three phases, from “the foundation,” in weeks one through four to “accompaniment mastery” in months three through six. 

When learning to play music, AI could certainly be used as a basic outlining tool or guide for lessons. The material it suggested I learn began with basic open chords, G major, C major, and D major. It then went on to suggest I learn strumming patterns, and add chords as the weeks progressed. 

The first and most obvious issue is that chat gpt lives up to its reputation of being overly flattering and unrealistic. Someone who can sing a little but struggles to play an instrument is likely going to need more than three to six months to be able to play well enough to accompany their singing. Sitting there reading the unnecessary flattery also wasted time that could have been spent focused on more detailed instruction. 

Despite its reputation for being able to generate things faster than anything else, chat gpt took three times as long to generate a chord diagram for the C chord than it took me to just find a website that somebody created with the same information. Chat gpt further failed to explain the fingering of the chord, while the site “Online guitar books dot com” owned and written by musician Genaaron Diamente from Melbourne, Australia, offered thar information right under the diagram.  

And just like with writing music, using AI to learn to play still leaves out those things only a human teacher, or at least humans supporting you and cheering you on as you teach yourself would provide, such as telling you what your playing sounds like and noticing whether you seem encouraged and determined or are just going through the motions and might not actually want to play. 

Experiment # 2: Phoenix Radio

As a local station that is both classic and contemporary, Phoenix Radio mixes old school R&B, Jazz, Blues, Rap, Hip-Hop, and Reggae with talk shows focusing on today’s issues. The station is continuously growing, reaching new audiences every week as we add new programs to our lineup and of course, new articles to our blog, all while allowing listeners to enjoy their favorite music from back in the day. 

AI cannot keep up with us. It has apparently never heard of us, and can only find us when told we exist. For this experiment, Microsoft Copilot was asked “Can you tell me about radio stations in Utica, New York?” It responded with a list of what appeared to be every radio station that ever existed in Utica, New York….except Phoenix Radio. 

When asked “What about Phoenix Radio?” Microsoft Copilot AI then appeared to find the very page you’re reading right now, and offered a brief, though slightly incorrect, summary of the station. It appeared to only be able to partially read the first part of the web page, as the only person listed was Scott Carr, the newest member of the Heat Squad, featured in the latest article on the main part of the page. The AI bot completely missed that the article was written by the radio station’s owner and SHE-e-o, Cassandra Harris-Lockwood, as it describes the site owned only by “Phoenix Radio, Inc, a community-focused media organization.” AI further failed to tell the difference between evergreen content on a web page and a comment section, using a fan’s mention of “DJ Butter” on the site “Online radio box”  to conclude that “DJ Butter” is a show. Anyone who has ever listened to the station knows that DJ Butter is a person.

Switching back over to chatgpt and asking “Who are the DJs of 95.5 The Heat: Phoenix Radio?” generated results that at least correctly identified DJ Butter as a person, but could only list two more people, DJ Tone and Bobby Bounce, and could not find the schedule or the names of any more members of the Heat Squad. When prompted with individual names, it could only read portions of the page or pick up articles about the founding of the station posted on the Utica Phoenix newspaper’s website in 2018. 

When looking for information about current events in music, or insight into the local music scene, AI is not only a poor source, it is something to avoid entirely. 

Experiment # 3: Some alternative facts about punk/alternative rock

Current and local events were certainly a failure, but there was still hope that AI could teach a bit of music appreciation and history. For this experiment, the prompt was “Teach me about the beginning of the punk and alternative scene in the 1980s.” 

While I am far from an expert on this subject, it is my favorite genre of music, so I would at least be able to tell if the bot just started spitting out the names of things that weren’t bands, or if it gave me a timeline off by an entire decade. It didn’t quite do that, but it didn’t provide anywhere near an accurate picture either. 

The biggest mistake “open ai,” which opened up a search engine on chatgpt, made in this experiment was to nearly leave out an entire city. According to the outline of the history of punk and alternative music generated by this chat gpt search, there was only one punk band to emerge out of Minneapolis, Husker Du. 

Husker Du was indeed an important punk band, they were just far from the only one in Minneapolis in the 1980s, a fact copilot can only seem to pick up when asked directly for a list of punk and alternative bands in Minneapolis in the 1980s. Following the links provided, and actually reading some of the articles that come up but aren’t picked up by AI teaches us that there were more than forty punk/alternative bands to come from Minneapolis, many of them from the same era. 

Unlike the material offered when asked for lessons, AI did not waste time with unnecessary flattery when asked about 1980’s punk and alternative history. And it didn’t get things completely wrong, like mistaking a person for a show, as it did when prompted to teach about current and local events in music. If you searched further into some of the bands the bot did mention, such as Black Flag and Joy Division, you would indeed find that these are important punk bands, and be encouraged to learn more. It just made a lot of mistakes by omission. If you knew nothing about the punk and alternative scene of the 1980s, relying on AI might leave you with a startlingly incomplete, and therefore inaccurate, picture. 

When learning about music, AI can serve as a helpful starting place or a source of misinformation. Searching using slightly different key words, typing the same ones in a different number of times, or even just performing these same searches on a different day may yield slightly or even entirely different results, perhaps better, maybe worse.  Overall, AI may give you a few ideas to take to your music teacher, favorite music expert, or local music personality, but it could never replace any of those people.