Artificial Intelligence (AI) can pick the hits!

[davideduardo]

Here is a report from StudyFinds at Simon Cowell replaced by a robot? Scientists use AI to pick hit songs

A robot could be coming for the job of prominent music producers and talent show judges like Simon Cowell, according to new research. Scientists have utilized artificial intelligence to identify hit pop songs with an impressive 97 percent accuracy. Such a computer system could render TV talent show judges redundant, replicating their skills at a significantly reduced cost.

It goes on to say,

With tens of thousands of songs released daily, it becomes challenging for apps like Spotify, Tidal, and Deezer to select which ones to add to playlists. Previous attempts to identify songs that will resonate with a large audience have had only a 50-percent success rate

(Free access site)

 

[TheBigA]

That's why you have a marketplace. Give people the choice to pick what they want.

The problem is AI really can't pick a future hit. Just like AI can't help me pick a winning stock.

It's all based on what worked in the past. The bad news is that popular taste doesn't always work that way.

 

[KeithE4]

That's why you have a marketplace. Give people the choice to pick what they want.

The problem is AI really can't pick a future hit. Just like AI can't help me pick a winning stock.

It's all based on what worked in the past. The bad news is that popular taste doesn't always work that way.

Nor can AI pick next year's Super Bowl, World Series, or Kentucky Derby winner. The events exist for a reason. Same goes for the recording industry, I would think. That would be up to the paying customers who like what they like, computers be damned. Also, trends can change quickly (see: The Beatles' 1963 failures in the US market. Along came 1964...).

It can probably make an educated guess, but as long as humans are creating and purchasing (or betting, in the sports context) content, I don't see how AI could make any difference whatsoever. Seems to me that anyone who would place money based on the predicted success of anything by a computer has a really good chance of losing.

 

[Mediafrog+]

The problem is AI really can't pick a future hit. Just like AI can't help me pick a winning stock.

It's all based on what worked in the past. The bad news is that popular taste doesn't always work that way.

Evolution of musical tastes and preferences occurs when people get tired of existing genres and are ready for something new.

I guess the question is: Can AI ever get bored? Weary? Yearn for change?

 

[tall_guy1]

A lot of mass appeal type music doesn't seem to resonate anymore (like "Jumpin'" by Pitbull) and some only resonate in certain areas/countries, so it would be impossible for it to choose something that would universally resonate (unless they maybe only go through a Taylor Swift catalog.)

 

[davideduardo]

That's why you have a marketplace. Give people the choice to pick what they want.

The problem is AI really can't pick a future hit. Just like AI can't help me pick a winning stock.

It's all based on what worked in the past. The bad news is that popular taste doesn't always work that way.

I think you are right. It can pick songs that sound like past hits... a lot like past hits. But it can't pick, to use an extreme example, "Don't Worry, Be Happy". Or, if you did not like that comparison, "Don't Let The Dogs Out". I could go on with "Tell Laura I Love Her" or "Teen Angel" but I'll stop now.

 

[davideduardo]

I guess the question is: Can AI ever get bored? Weary? Yearn for change?

Or interpret the emotion of lyrics, style and orchestration?

 

[davideduardo]

Also, trends can change quickly (see: The Beatles' 1963 failures in the US market. Along came 1964...).

But that was due to the first releases being on indies and not vehemently promoted at the $tation level.

It can probably make an educated guess, but as long as humans are creating and purchasing (or betting, in the sports context) content, I don't see how AI could make any difference whatsoever.

I think it would have trouble with predictions based on change and novelty value. How could it have, back in the day, predicted "hittiness" for The Girl from Ipanema?

Seems to me that anyone who would place money based on the predicted success of anything by a computer has a really good chance of losing.

Or missing the best opportunities.

 

[michael hagerty]

Okay, so a careful read of the article shows AI isn't really "predicting the hits", it's choosing songs that make humans' brains light up in ways that indicate they like those songs---which "enabled the team to predict market outcomes".

This is a Rube Goldberg path to the oldest thing in the book---the turntable hit. Because, as a chunk of us here on this board know, there's a big difference between them liking it and them buying it.

 

[Kelly A]

Evolution of musical tastes and preferences occurs when people get tired of existing genres and are ready for something new.

I guess the question is: Can AI ever get bored? Weary? Yearn for change?

AI isn't sentient, so it can't get bored.

 

[Mikey Radio]

It's been interesting to read some of the takes on AI in general. I saw a news article just this morning advising young people not to go to law school or enter the legal field, as those jobs would soon be greatly reduced and in some cases taken over by AI, which could write legal briefings faster, better and much cheaper than human lawyers. The same article claimed that the best protection from AI taking your job was to find work in blue collar trades. Of course, conversely, one of the biggest "lessons learned" by states in the 'rust belt' of the USA is that unskilled blue collar jobs don't pay the bills as they once did, and many of those jobs in mills, heavy manufacturing and the steel industry went overseas or to other countries decades ago.

 

[TheBigA]

Of course, conversely, one of the biggest "lessons learned" by states in the 'rust belt' of the USA is that unskilled blue collar jobs don't pay the bills as they once did, and many of those jobs in mills, heavy manufacturing and the steel industry have gone overseas or to other countries.

There's always a cheaper way to do something. When the motivation is profit, you're always looking to cut expenses. So one day you're celebrating getting that big raise, and then a couple years later (when the contract is up), you get cut because you're too expensive. One of my friends is a striking Hollywood writer. He tells me they know the result will be fewer writers, because they're getting to expensive. But it's the principle of the thing.

 

[davideduardo]

Okay, so a careful read of the article shows AI isn't really "predicting the hits", it's choosing songs that make humans' brains light up in ways that indicate they like those songs---which "enabled the team to predict market outcomes".

This is a Rube Goldberg path to the oldest thing in the book---the turntable hit. Because, as a chunk of us here on this board know, there's a big difference between them liking it and them buying it.

And that is something that record promoters and, even, the general public, does not get: a song may sound “nice” on first listen, but listeners don’t establish it as a real deep-seated like… or dislike… until somewhere around the 5th to 7th full exposure. those of us who have done thousands of music tests and, maybe, a million call-out interviews see how songs progress after one, two, three and more exposures.

One of the researchers who worked with me observed that this is like falling in love. “Love at first sight” is rare and it takes repeated exposure to deepen the relationship.

 

[landtuna]

Okay, so a careful read of the article shows AI isn't really "predicting the hits", it's choosing songs that make humans' brains light up in ways that indicate they like those songs---which "enabled the team to predict market outcomes".

What's the difference between "lighting up a potential hit" and "light up to push the pre-set". Same part of the brain or different?

 

[michael hagerty]

What's the difference between "lighting up a potential hit" and "light up to push the pre-set". Same part of the brain or different?

I think it's the same part, and it either lights up (potential hit) or doesn't (push the pre-set).

 

[AMRadioGuy]

Nor can AI pick next year's Super Bowl, World Series, or Kentucky Derby winner. The events exist for a reason. Same goes for the recording industry, I would think. That would be up to the paying customers who like what they like, computers be damned. Also, trends can change quickly (see: The Beatles' 1963 failures in the US market. Along came 1964...).

It can probably make an educated guess, but as long as humans are creating and purchasing (or betting, in the sports context) content, I don't see how AI could make any difference whatsoever. Seems to me that anyone who would place money based on the predicted success of anything by a computer has a really good chance of losing.

Of course, promotion certainly helps. Part of why it took until '64 for the Beatles to catch on here in the states is because the head A&R guy at Capitol, Dave Dexter, was not a fan of the group so a lot of their singles came out on small regional labels.

So that's another variable that AI would certainly need to take into consideration.

 

[AMRadioGuy]

And that is something that record promoters and, even, the general public, does not get: a song may sound “nice” on first listen, but listeners don’t establish it as a real deep-seated like… or dislike… until somewhere around the 5th to 7th full exposure. those of us who have done thousands of music tests and, maybe, a million call-out interviews see how songs progress after one, two, three and more exposures.

One of the researchers who worked with me observed that this is like falling in love. “Love at first sight” is rare and it takes repeated exposure to deepen the relationship.

That's pretty fascinating, David!

 

[Michael Rogers]

Interesting to read the different comments on AI and "picking the future hits". I worked in radio 1984-1994 and was informed in a staff meeting that synthetic voice was going to "replace us all". Perhaps technology is getting closer to perfection, but I still note from most of the platforms/services I have perused, an occasional flub with intonation and even cringe-worthy results.

I am entering radio now, simply as a hobby and with evangelical motives. I hope to be issued an LPFM license in the near future...I don't NEED to make a buck... I'm not wealthy, but I AM rich. Any kernels for wisdom would be appreciated. I am looking for playout software. Reviews of BUTT appeared to make it a "top choice", but Reddit conversations have dissuaded me so far. Any suggestions?

I intend to stream the audio, as well as utilize FCC authorized frequency when the opportunity arises. Oh.... "Michael Rogers" is my former on-air name when I was a small market DJ.

 

[Miss Tuned]

Reviews of BUTT appeared to make it a "top choice", but Reddit conversations have dissuaded me so far. Any suggestions?

The only BUTT I'm aware of in this context is a streaming package called Broadcast Using This Tool, which takes an audio input and turns it into a stream for Icecast or Shoutcast servers. It used to do that simple job just fine when it was my job to care about this stuff (and I'm glad to see it's still active and updated) but I'm not sure of its relevance to this thread. In any case, it's not a playout package, it is just a simple tool.

 

[michael hagerty]

I'm sorry---AI? BUTT?