you going back two years is whip lash, jesus. i trained a 2 hour dataset for 40 hours.
Oh, you mean it took your system 40 hours to process the data? That's not usually how this stuff is reported, but ok. (If you had a computer that was twice as fast, it would take half as long)
Yeah, training on the whole MJ catalogue should definitely be the goal. I dunno the exact length of every song, but it's gotta be about 20 hours of music. And as I said, speech and live singing could be additional data.
I'm currently reading a book called 'Mood Machine: The Rise of Spotify and the Cost of the Perfect Playlist'.
I've heard of the book.
Yeah, those playlists are just there to increase Spotify's profits (ie by paying less money to artists). It's just a scam, that's all. But that's nothing to do with AI or the quality of AI songs.
How much cheaper can it get? The royalty situation with streaming is already deplorable.
Free, obviously.
Maybe royalties aren't the way forward. What next, if you hang a door in somebody's house, you get to charge them a dollar every time they go through it?
Who's going to pay the artists?
It doesn't matter?
Music has been around since before the concept of money was invented. There will always be people who make music, whether they get paid or not. You're focusing on the wrong thing.
Leaving aside TS and Jay-Z, most musicians won't end up as billionaires, that's true. But Daniel Ek, Spotify CEO, is a billionaire.
That's just the business model of a company that wants to make money from music. Again, you're focusing on
the music industry. Don't do that. Focus on
music. It doesn't matter if the industry dies. Music is what's important.
What? Human beings invented AI only few years ago. How could you do that in the end of 90s
Um, no? AI has been around since about the 1950s, it's not new, lol. If you look hard enough you can find references to artificial neural networks going back to 1795.
Again, in 1999, as part of a masters thesis, when I was 22, I programmed a system (actually 3 systems) that could study a pattern, learn the difference between different categories, and categorize new data accordingly to what it had learnt.
No way. Computers appeared only in 1990-2000s
I got my first computer in 1982.
Please look it up. Even using a Google AI search will tell you computers were around in 1837!
So glad that it's much faster today. One tap and it's sent in a split second.
I'm not, lol. Making it so quick/easy led to waves of spam so powerful they forever ruined the entire system. Email today is unusable. I never touch it outside of work.