How is Phase handled?

Official support for:
2359 posts since 24 Nov, 2012

Post Tue Feb 19, 2019 2:10 am

A_SN wrote:
Mon Feb 18, 2019 11:26 pm
Well I don't want to go on a whole rant about how we've reached peak hype for AI/ML with little of actual practical value to show for it mostly in artistic fields (generating some weird stuff automatically from some input doesn't rank very high for me), so I'll just tell you that I have no interest in neural voodoo synthesis. I really don't see the interest, none of what I've seen or heard about it seems compelling, but more importantly it's the approach that I find uninteresting due to the lack of control. Again it's largely voodoo, you throw a bunch of "training" and massive amounts of computing power at something and it makes a weird poor quality sound, why? I'm much more interested in creating rational tools to let human intelligences control every aspect of creation or transformation. You'd think that given what I do I'd be interested in whatever that can create weird sounds but actually I'm not, my interest was always to control sound in a direct and clear way based on sound principles, not just throwing algorithms at some input to see what happens. I'm smart enough to do what you want but more importantly I'm smart enough not to do it

I think it's perhaps a philosophy thing, I see computers as fancy calculators that can do pixels and sound samples and I know exactly what I want to do. It seems that AI enthusiasts will be happy with any results they consider cool enough (they set the bar for cool rather low) and see computers as potentially magic and try to make them do unscientific magical things that are easy to be made to work poorly but impossible to be made to work well due to the usually chimeric nature of the desired goal (e.g. "automatically colourise a black and white image with no other input", easy to make it guess that the sky is blue and grass is green, impossible to make it generally close to accurate, enthusiasts usually dismiss such fundamental limitations by extrapolating that everything will get better as we approach the techno-rapture and their AI god kicks the abrahamic god in the teeth). I'm trying not to rant too much, but isn't it amazing how so much of computer """science""" in the late 2010s is mostly becoming "hey let's just throw TensorFlow and a bunch of GPU cores at it and see what happens" and they can only speculate as to what their algorithms actually do? And a lot of the people involved firmly believe in the AI techno-rapture they call "The Singularity" and that their work will help immanentise the AI eschaton. It's like a philosophical degenerative disease that makes potentially very capable people waste their time on pointless endeavours, kind of like string theory did for astrophysics, but more as a new religion for atheists. Ah well, that leaves more low hanging fruits for me to pick.

couldn't agree more - and I have an academic background in some of this.
what you don't know only makes you stronger

11 posts since 24 Jan, 2019

Post Tue Feb 19, 2019 5:27 am

You're not going to get out of the realm of shrieks until you get AI to assist with shaping the drawings/waves.

996 posts since 6 May, 2008 from Poland

Post Tue Feb 19, 2019 9:52 am

woggle wrote:
Tue Feb 19, 2019 2:10 am
couldn't agree more - and I have an academic background in some of this.
:tu: I didn't expect anyone to agree with me, glad I'm not the only one who sees things like this. If I wasn't a developer I could complain about the way things are going but since I feel that all this clears the competition by keeping them busy exploring dead ends I feel rather gleeful about it. And it makes companies more interested in investing in/researching GPGPU stuff which is good for me.
pittsburghjoe wrote:
Tue Feb 19, 2019 5:27 am
You're not going to get out of the realm of shrieks until you get AI to assist with shaping the drawings/waves.
Nonsense. Besides so far it's just you (other people are more interested in more practical missing features like stereo support or importing) and it's all because of peak hype. Ten years from now when people stop believing that throwing "AI" at everything is what the march of time requires of every developer then no one will bother me with their non-solutions in search of a problem to solve.


Again, don't shriek at me to create the projects you want, do it yourself, you know what you want and it probably takes minimal skills and effort to write the necessary glue code. No one is going to make me waste my time doing things that I consider pointless just because they want to see what happens.
Developer of Photosounder (a spectral editor/synth), SplineEQ and Spiral

11 posts since 24 Jan, 2019

Post Thu May 09, 2019 5:01 pm

Return to “Photosounder”