At its best, artificial intelligence can assist people in analyzing data, automating tasks and developing solutions to big problems: fighting cancer, hunger, poverty and climate change. At its worst, AI can assist people in exploiting other humans, damaging the environment, taking away jobs and eventually making ourselves lazy and less innovative.<\/p>\n
Likewise, AI is both a boon and a bane for the music industry. As a recording engineer<\/a> and professor of music technology and production<\/a>, I see a large gray area in between.<\/p>\n The National Academy of Recording Arts and Sciences has taken steps to address AI in recognizing contributions and protecting creators. Specifically, the academy says, only humans are eligible<\/a> for a Grammy Award: \u201cA work that contains no human authorship is not eligible in any categories.\u201d <\/p>\n The academy says that the human component must be meaningful and significant to the work submitted for consideration. Right now, that means that it\u2019s OK for me to use what\u2019s marketed as an AI feature in a software product to standardize volume levels or organize a large group of files in my sample library. These tools help me to work faster in my digital audio workstation. <\/p>\n However, it is not OK in terms of Grammy consideration for me to use an AI music service to generate a song that combines the style of say, a popular male folk country artist \u2013 someone like Tyler Childers \u2013 and say, a popular female eclectic pop artist \u2013 someone like Lady Gaga \u2013 singing a duet about \u201cStar Trek.\u201d<\/p>\n<\/p>\n This song, one of the most popular on Spotify in Sweden, was banned from the country\u2019s music charts after reporters discovered that it was substantially generated by AI.<\/p>\n The gray zone<\/p>\n It gets trickier when you go deeper. <\/p>\n There is quite a bit of gray area between generating a song with text prompts and using a tool to organize your data. Is it OK by National Academy of Recording Arts and Sciences Grammy standards to use an AI music generator to add backing vocals to a song I wrote and recorded with humans? Almost certainly. The same holds true if someone uses a feature in a digital audio workstation to add variety and \u201cswing\u201d to a drum pattern while producing a song.<\/p>\n What about using an AI tool to generate a melody and lyrics that become the hook of the song? Right now, a musician or nonmusician could use an AI tool to generate a chorus for a song with the following information:<\/p>\n \u201cWrite an eight measure hook for a pop song that is in the key of G major and 120 beats per minute. The hook should consist of a catchy melody and lyrics that are memorable and easily repeatable. The topic shall be on the triumph of the human spirit in the face of adversity.\u201d<\/p>\n If I take what an AI tool generates based on that prompt, write a couple of verses and bridge<\/a> to fit with it, then have humans play the whole thing, is that still a meaningful and significant human contribution? <\/p>\n The performance most certainly is, but what about the writing of the song? If AI generates the catchy part first, does that mean it is ultimately responsible for the other sections created by a human? Is the human who is feeding those prompts making a meaningful contribution to the creation of the music you end up hearing? <\/p>\n AI music is here<\/p>\n The Recording Academy is doing its best right now to recognize and address these challenges with technology that is evolving so quickly. <\/p>\n Not so long ago, pitch correction software like Auto-Tune caused quite a bit of controversy<\/a> in music. Now, the use of Auto-Tune, Melodyne and other pitch correction software is heard in almost every genre of music \u2013 and no barrier to winning a Grammy.<\/p>\n Maybe the average music listener won\u2019t bat an eye in 10 years when they discover AI had been used to create a song they love. There are already folks listening to AI-generated music by choice today.<\/p>\n You are almost certainly encountering AI-generated articles<\/a> (no, not this one). You are probably seeing a lot of AI slop<\/a> if you are an avid social media consumer. <\/p>\n The truth is you might already be listening to AI-generated music, too. Some major streaming services, like Spotify, aren\u2019t doing much<\/a> to identify or limit AI-generated music on their platforms. <\/p>\n On Spotify, an AI \u201cartist\u201d by the name of Aventhis<\/a> currently has over 1 million monthly listeners and no disclosure that it is AI-generated<\/a>. YouTube comments on the Aventhis song, \u201cMercy on My Grave<\/a>,\u201d suggest that the majority of commenters believe a human wrote it. This leads to questions about why this information is not disclosed by Spotify or YouTube aside from \u201c[h]arnessing the creative power of AI as part of his artistic process\u201d in the description of the artist. <\/p>\n<\/p>\n This AI-generated song has millions of listens on Spotify and views on YouTube.<\/p>\n AI can not only be used to create a song, but AI bots can be used to generate clicks and listens<\/a> for it, too. This raises the possibility that the streaming services\u2019 recommendation algorithms are being trained to push this music to human subscribers. For the record, Spotify and most streaming services say they don\u2019t support this practice. <\/p>\n Trying to keep it real<\/p>\n If you feel that AI in music hurts human creators and makes the world less-than-a-better place, you have options for avoiding it. Determining whether a song is AI-written<\/a> is possible though not foolproof. You can also find services that aim to limit AI in music.<\/p>\n