[angerman]

This a good question. But this hits a very interesting point. Let us discard crappy songs for a bit. And focus on marketing and other reasons. Marketing plays an important role. Yet there are artists who cannot afford the marketing or (and this is very sad in my opinion), the produced a piece that was ahead of it's time or just didn't fit the current market trends.

These kind of services allow us to explore the full spectrum of creativity, instead of homing in on current market trends. If we do not support the outlier (or in this case leave them lying by the sidelines and ignore them), we loose a part of cultural diversity. Not everyone has the same taste, but how do you experience new things when you just follow the herd?

Coincidentally I was talking with someone about House of Cards over lunch. And whether this is correct or not, it was claimed that House of Cards was the results of machine learning and figuring out that the majority wanted Kevin Spacey and a political drama. Now if this is the future of tv, this will lead inevitably to less diversity. This would be a very sad development.

This is precisely why I think that services like this one are very important to keep us culturally diverse. (at least in the limited spectrum of songs available on spotify in this case.)

[mseebach]

> it was claimed that House of Cards was the results of machine learning and figuring out that the majority wanted Kevin Spacey and a political drama. Now if this is the future of tv, this will lead inevitably to less diversity.

If the algorithms are actually any good, they should lead to more, not less, diversity. The algorithms would be able to detect that enough people want to watch a Kung-Fu thriller set in medieval South India starring Summer Glau to be viable - the point of these things is that "big data" machine learning can process several orders of magnitude more data than a Hollywood-bubble studio executive gut feeling ever could.

Formulaic mass-market blockbuster stuff is the product of humans and their gut feelings, and broad averages over the mass market is the highest abstraction they can handle. Algorithms have no such limitations.

The truly niche/innovative out-of-left-field stuff exists because it was able to bypass concerns about mass-market viability - if anything, the algorithms will pick up on niche trends much quicker, bringing it to a larger audience.

[DanBC]

> If the algorithms are actually any good, they should lead to more, not less, diversity. The algorithms would be able to detect that enough people want to watch a Kung-Fu thriller set in medieval South India starring Summer Glau to be viable

They don't want "viable", they want blockbuster.

When a film in a series makes only half a billion it's seen as a failure and people are called in to make the next one a success.

http://uk.ign.com/articles/2015/05/19/simon-pegg-says-origin...

> “They had a script for Star Trek that wasn’t really working for them,” [Simon Pegg] told magazine Radio Times (via The Guardian). “I think the studio was worried that it might have been a little too Star Trek-y.

> “Avengers Assemble, which is a pretty nerdy, comic-book, supposedly niche thing, made $1.5bn. Star Trek: Into Darkness made half a billion, which is still brilliant. But it means that, according to the studio, there’s still $1bn worth of box office that don’t go and see Star Trek. And they want to know why.”

> He further explained “People don’t see it being a fun, brightly coloured, Saturday night entertainment like the Avengers,” adding that the solution was to “make a Western or a thriller or a heist movie, then populate that with Star Trek characters so it’s more inclusive to an audience that might be a little bit reticent.”

[mseebach]

> They don't want "viable", they want blockbuster.

That may be what they want, but that's not all they make. To wit, plenty of shows and films (indeed, a large majority in number) aren't blockbusters, and don't appear to have been intended as such.

Films that are expected to make 1.5bn, but only make 0.5bn are failures in that regard, sure - but that doesn't mean that there isn't a large market for films that "only" make 0.5bn.

[jarman]

>They don't want "viable", they want blockbuster. They want predictable profit. If machine learning will allow producers to make better predictions on niche market - they will go there too

[stkni]

>> Now if this is the future of tv, this will lead inevitably to less diversity. This would be a very sad development.

Debatable, surely. Especially since House of cards on NetFlix was a remake of a BBC drama from 1990 [1]. Suggesting that diversity exists because someone could be bothered to make it, presumably without data-mining. The point is that this diversity is, predictably, not in the mainstream.

[1] http://en.wikipedia.org/wiki/House_of_Cards_(UK_TV_series)

[jcurbo]

Which was originally based on a book!