50 buyins is pretty reasonable for an average (which means losing because of the rake) player for a year. You can plug in the numbers into one of the poker variance calculators. At 1 hour per day (60 hands), 10 BB/100 rake, NLHE 6-max cash games only, you get the 95% confidence interval of -5150 BB to 770 BB after 1 year.
@zone411
From 8/13/2010, 12:49:03 PM till now, @zone411 has achieved 1884 Karma Points with the contribution count of 433.
Recent @zone411 Activity
Fun idea. We're about test out melodies I've created together with a generative neural net and see how they're rated computer to real melodies. The plan is use Amazon's Mechanical Turk but one problem is what to do about real melodies that people will mark as already familiar to them. I think making it a comparison with unfamiliar melodies should be fine?
The most of obvious way would be to use the same test as previously and just compare the scores.
Not really any new info there for anybody who has followed basic economic news but still I would take these sorts of predictions much more seriously if he offered some kind of proof that he's putting his money where his mouth is. If he thinks the dollar will collapse, he could post a proof that he's got some crypto wallet with his net worth in it or some S&P500 puts, whatever. Otherwise, it's just more empty Twitter chatter.
Covid won't help: https://www.thelancet.com/journals/eclinm/article/PIIS2589-5....
America Reached One Million Covid Deaths
5 points • 0 comments
I believe the real AI music revolution is still ahead of us. Most existing tools don't really use modern deep neural nets and the ones that have just been released and do are quite good (e.g. Synthesizer V). The project I released yesterday uses all research advances (such as diffusion models) and more (https://www.youtube.com/playlist?list=PLoCzMRqh5SkFwkumE578Y...) and I think it's a large leap forward over what the existing melody generation software can do.
I think that pretty soon we'll see many hit songs that have a large contribution from AI models. It's possible that the variety can increase as well and new styles can appear, as the "good" generated songs are fed back as training data.
I found that it's best to avoid most of these webpage explanations that pop up in Google (often on Towards Data Science and Medium). You can get a better understanding by reading intro sections of actual research papers.
Yes, I am. It does many things well but last time I used it (about a year ago), I couldn't get it to do what I focused on with this project: create a catchy melodic hook that could be a base of a hit song or an earworm commercial jingle. I believe this requires an element of surprise in a melody - it's much easier to just create something pleasant-sounding, without anything fancy related to neural nets.
Thanks! I wasn't familiar with a couple of these, I will definitely check them out.
Thanks. I used Piano VFX (it's a bit buggy but workable) and VSDC (it's missing any sort of more advanced functions like batch rendering but does visualizations very well).
I'm guessing the link comes from my Show HN post earlier today about a generative diffusion process neural net assistant which let me create 60 pretty catchy melodies that linked to this page (https://news.ycombinator.com/item?id=31343944).
One interesting question will be about copyrights. There are other melody-making programs but I'm not aware of anything catchy or worth registering with the U.S. Copyright Office being created by them. It costs $8.50/song. I've registered all 60 melodies I've posted so far since they all required my creative input, but with further advancements, the creative input will be reduced to cherry-picking best melodies. Eventually, a high enough percentage of the melodies created will be good enough that somebody will attempt to just register all of them. It's unclear whether they'll be judged as worthy of copyright protection under current laws or if these laws will be changed.
Another very interesting question is what's the number of "good" melodies. We could define this as being rated by random people at a level equivalent to an existing melody from a database (e.g. Hooktheory's) that they're not familiar with.
> In a sense, the lyrics could be spoken and then applied to the .midi melody. The result would sound similar to the SoundCloud vocal OP link, but with "words".
Yes, Synthesizer V Studio can do it quite well! I've added some silly lyrics to a few of these melodies and I was quite happy with the results. I'll upload them to the same YouTube channel. Here is a cover song that somebody made: https://www.youtube.com/watch?v=ettSVy4K_DY.
Not yet. The current version is not fully automated and it requires human input, so it will take a decent amount of work and documentation to make it ready for users. It also mixes C++ with Python in a few places for performance reasons when generating melodies and I don't know if that's even possible on Colab?
Show HN: Catchy melodies made with a diffusion-based neural net assistant
37 points • 14 comments
Space Time is excellent compared to Hossenfelder's videos.
She makes videos about many topics on which she's not even close to an expert.
Her Simulation Theory video was absolutely awful (https://lech.substack.com/p/sabine-hossenfelders-video-the-s...), so it's not just a gut reaction.
Agree. I stopped watching her stuff after her really terrible video about the Simulation Hypothesis, which showed that she didn't even grasp the argument. Since she feels free to opine about things she clearly knows nothing about, I can't know if she's also talking about other topics without any real understanding (Gell-Mann Amnesia).
Her video on the Simulation Hypothesis was very poor https://lech.substack.com/p/sabine-hossenfelders-video-the-s....
site design / logo © 2022 Box Piper