If you add cameras with vision processing and AI, they will read the room better than a human
What makes you cringe in tango?
- Thread starter TTanguero
- Start date
brujo
Well-Known Member
One thing that AI does that is interesting is that sometimes it lowers the barrier to some stuff that could be done, but used to be super expensive or require a lot of expertise.I am not touching AI generated music
I volunteer with a bunch of kids that absolutely hate writing from a low income setting, and we came up with an activity where they could write a song and we would use Suno to make them into actual songs. I never seen them so engaged in an activity, to the extent that we had to repeat it twice because they liked it so much. They were elementary kids so the songs were hilarious.
Tons of fun things to do with AI music too. Make songs more accessible to non Spanish speakers. So for example taking a sad popular English song, pasting the lyrics and asking it to recreate it as a tango. Or taking the lyrics to famous tango songs like antiguo reloj de cobre and translating it to Chinese and making it mandapop. Also when learning the difference between tango bands trying hard to describe to Suno what Pugliese would sound like. Or even imagine what a tango sound with Mongolian throat singing would be like. It’s a fun creative medium.
opendoor
Forum Master
Exactly thats why I started collecting shellacs. I want to be a minority. I can do things that only a few still can. The ones I know of are Zingo in Brussels, Michael Rühl in Berlin, Michael Thiele in Cologne, Thomas Jordan in Dresden, Volkmar Hess in Viersen and Fernando Vicente of TT.
What I meant was I am not going to open can of worms about AI generated music. People get too emotional. If you look back in history, initially there is opposition to new tool which makes creating art easier. Then a new class of artists emerge who create arts using new tools, which previous generation looked down upon. In music creation you can see parallel. Think of how electronic music synthesizer changed the process of creating music.
I think in future musicians will use AI to create music. It is inevitable. They might use AI for discovering or prompting choices in rhythms, melody, tune, etc from which they can pick to develop finished music. Or they can use AI to generate finished music. People will find hard to distinguish between AI generated music and human generated. I have read at least one comment or confession by someone - she liked the music but after it was told that is was AI generated, she couldn’t admire it.
For the musicians, AI can bring down costs of music production and its distribution.
Like I said it is opening a can of worms.
pianistamediano
Forum Master
I've just got a new laptop with 64GB of RAM and the laptop version of the 5080 to play with, so I'm planning to run as many local models as I can and have them interact with each other in fun ways.
AI as a tool is great IF you know what you're doing, which most people don't, hence the explosion of sloppy AI slop and tasteless generated content. I use it to prototype a lot of boilerplate code and tedious work, as well as turn pseudocode for algorithms into actual code — it's surprisingly good even when the algorithm is non-trivial, but I still have to check every line and write extensive tests (which I do even for my own work) to make sure I don't miss anything. Even so, when it comes to proprietary codebases, it doesn't know about the libraries of custom functions and methods, nor the overall structure, so any code it provides is isolated.
Vibe coding is good and all at breaking down barriers for, say, app development, but you often run into severe security/authentication issues and scaling them up is an exercise in hilarious failure.
In my opinion, most people are ignorant because they have little to no idea nor understanding about how AI works. Their opinions are essentially worthless when it comes to what AI is and its capabilities, present or future. It's a free-for-all when discussing the consequences or ethical concerns, of course.
I've often thought about writing and training a model to come up with tandas, but I still need to clean, cross-reference and verify the metadata for my whole library, which is about 90+ GB, so... Maybe in the next century or so.
AI as a tool is great IF you know what you're doing, which most people don't, hence the explosion of sloppy AI slop and tasteless generated content. I use it to prototype a lot of boilerplate code and tedious work, as well as turn pseudocode for algorithms into actual code — it's surprisingly good even when the algorithm is non-trivial, but I still have to check every line and write extensive tests (which I do even for my own work) to make sure I don't miss anything. Even so, when it comes to proprietary codebases, it doesn't know about the libraries of custom functions and methods, nor the overall structure, so any code it provides is isolated.
Vibe coding is good and all at breaking down barriers for, say, app development, but you often run into severe security/authentication issues and scaling them up is an exercise in hilarious failure.
In my opinion, most people are ignorant because they have little to no idea nor understanding about how AI works. Their opinions are essentially worthless when it comes to what AI is and its capabilities, present or future. It's a free-for-all when discussing the consequences or ethical concerns, of course.
I've often thought about writing and training a model to come up with tandas, but I still need to clean, cross-reference and verify the metadata for my whole library, which is about 90+ GB, so... Maybe in the next century or so.
Whilst I agree with all that for many purposes where AI is considered, I think @Bailango touches on another area of use, irrespective of how poor 'AI' may be considered in that area at the moment.
I don't necessarily have to understand the physics of sound production to play an instrument. Vibe composing music (with AI) may be useful, or maybe not if you already have that musical skill. But maybe it assists some prospective amateur along the way. Maybe it allows later human refinement.
Lots of 'maybes'! I see a use for AI as a tool in several areas given that it has the information on which to draw. At the moment I don't see it writing a great novel, composing great music, or an artistic masterpiece, or even great video. It's a novelty in those areas (like fonts were with early printers). It's another tool to evaluate, to take from the toolbox, and use in the appropriate place - when we stop seeing it as a novelty and understand where it can be used best. There will be many mistakes before then!
And there will still be poor playlist choices because the poor consequences weren't reviewed and captured when someone stored their playlist. Neither were the choice 'rules' understood nor captured. But it might be a better proposal than someone's choices because not every milonga will hve a great DJ.
pianistamediano
Forum Master
I'd probably design it as a two-step model/process; one to compose individual tandas, another to figure out the "best" time to play it. I'd rather not mess around with computer vision and facial/body language analysis — there's issues with continually analysing people for the duration of the milonga, even if it isn't recorded.
brujo
Well-Known Member
One pitfall of being a technical person is that we often overthink the level of customization and data we think it is needed. If you have no DJs in a community, a Spotify tango playlist that is AI generated might be just good enough.
In the same way, the AI tooling can just be assistive rather than a full replacement. Gemini is smart enough to give you a suggestion for four possible artists to play in the next tanda after a Pugliese set, for example. Or it’s super easy to build an instagram AI assistant that can rattle off a list of neotango songs to put into a 50-50 traditional / alt set.
opendoor
Forum Master
It's strange that there are communities without DJs. We're all getting in each other's way, here. I find there are far too many tango DJs and far too many tango teachers around.
pianistamediano
Forum Master
That's for myself, not for the uninitiated.
Spotify already does that; if I wanted to design an idiot-proof model for people to use, I'd just write a basic GUI over a simplified version and restrict any access to the back end.
so I'm planning to run as many local models as I can and have them interact with each other in fun ways.
AI as a tool is great IF you know what you're doing, which most people don't, hence the explosion of sloppy AI slop and tasteless generated content.
Rather than make this about AI nuances and AI discussion with dance and music taking a back seat, my short comment is that using current generation of AI tool sets - a perfect analogy is understanding car engine to have a good car that drives well.
Those who understand under the hood are fine tuning the engine to make car run much better. A mass product car (AI) is below average compared to what you can do with a fine tuned AI for specific purpose. We are not at a stage where generic or AI for masses is at above average level. You need to know how things work to make best use of it.
The generic AI will get better and better at doing multiple things in the longer run. Fine tuning or customizing your own models will give only marginal gains. That days isn’t here yet.
Artist is likely a car driver. Once car is good enough, they will be able to drive it beautifully and smoothly. Without having to constantly check for whether engine is getting hot, radiator needs coolant, mileage is low, etc.
P.S. AI slop is inevitable. When YouTube enabled user generated content to be published for anyone to watch, it was all associated with cat videos. Lowering barriers means quality is the first victim.
pianistamediano
Forum Master
But cat videos are the peak of quality! I'd rather organisers use cat videos/GIFs/images to advertise their events than AI slop.
brujo
Well-Known Member
A friend of mine was promoting an event with a bar that unfortunately loved terrible AI posters. Older owner, it was impossible to convince them that they could do things like remove extra arms in the AI poster. Must have been their son or nephew that made them.
You kinda discover fairly quickly that the online vocal crowd that bitches about AI slop and screams 'this poster is AI so the event must be garbage' is not going to attend the event anyway. Normal people tend to be a little more forgiving and understanding of the stuff promoters have to go through to get events out.
You are comparing 20 years of progress in cat videos with only recent AI slop ?
With little attention, AI generation of multimedia is getting good. It will continue to improve rapidly. Yeah fake videos and pics will be a problem (just like all the slop created by people on social media over last two decades). I have seen some very interesting and creative videos and pics generated by AI. Those have been intentionally done by artists. They often use more than one AI tool in their workflow. I think in a few years we could see interesting output by average people who use AI. They could be generating educational content, entertainment content, cooking shows, etc. Today on YouTube you get very good content created in all kinds of category that is very useful. From learning about particle physics to multiple recipes for the same dish. Instead of having to film yourself, then edit, etc, if I can prompt AI with recipe, it will generate full video with voice and realistic kitchen. You can decide whether to show how to chop onions finely or simply show chopped onions, you can show multiple options. Today you have to film and record yourself to create all the footage which then you will edit. I can easily see 70% of time saved.