@rhempel I've always felt that being an artist does not give you a free pass (and it's one of the reasons why I always struggled to call myself an artist).

Even if the viewer doesn't see a discernible difference, I think considering energy consumption and algorithmic efficiency as (some of the) hidden costs of art making & viewing always should be a consideration. I also think it's a sign of maturity, both personal and in the larger picture as discipline, to be considerate of these things, even more so for animated or time-based pieces which are running nonstop at galleries/exhibitions/homes for weeks/months. Unfortunately, this is an extreme minority opinion in this field...

But energy aspects aside, especially for animated and/or interactive works, achieving a minimum frame rate (at the very least 30fps, better 60fps or 120fps) is important (not always though!) and failing to do so, can have a negative impact on the overall aesthetics, feel, experience, immersion etc. Optimizing a piece from 2-3 fps to 30+ fps can be a literal make-or-break switch...

Also for realtime audio, which is much less forgiving, hard time limits are enforced by the underlying OS/hardware audio driver. All (sonic) hell breaks loose if you don't manage to compute N new samples for the next time window in a shorter time than that window's duration (usually just a few milliseconds a slice)...

So back to your question: Is there a way to convey a elegance and design in implementing of algorithmic art? I think the answer here is: Yes. At least for the group of time-based works, these aspects are partially/indirectly conveyed precisely via working smoothly within these time & resource constraints, some of them enforced by our own perception thresholds, others by hardware. An animated or interactive system which feels sluggish hardly ever communicates "elegance" (where it matters and in an abstract sense, but even "glitch art" often works with a more or less artist controlled form of glitchiness). For interaction it can even very quickly become actively frustrating once certain action-response delays are exceeded regularly...

Some artists are trying to free themselves from the efforts having to deal with these things via means of clever framing and post-rationalization, i.e. by stating the artwork presented is merely a 1:1 instance of the actual artwork which is just a sequence of instructions given on paper (or as text file, as voice instruction etc.) and therefore has neither conceptual nor technical room for addressing these aspects... I can think of some cases where any form of optimization truly isn't desired (e.g. those concepts which communicate exponentially increasing complexity or time effects of a given system), but I'd say these are not the norm... Mostly it feels like: Purity for purity's sake and an easy way out... 🤷‍♂️

In some sense the problem for algorithmic artists is similar to that of game devs: Developing your idea on high-end hardware with top-range compute can be a recipe for disaster when deploying your artwork on lower spec hardware. In that sense a well optimized implementation or one which is aimed at lower compute resources, simply means the piece is much more adaptable and also generally less resource/energy intensive, something which we as society are maybe just not yet valuing sufficiently enough, but which I believe we will come to (value)... People will still want to have engaging digital art in the future, even if compute availability will be much more unevenly available/affordable, less abundant, and more artists, especially them, will be working/showing on devices with lower compute (relatively speaking) due to exploding hardware costs and centralization...

That quote I mentioned in the linked posts above is a perfect summary of this all:

"Too often in this industry hardware is used to solve software problems."

#AlgorithmicArt #InteractiveArt #TimebasedArt #Optimization #Performance #FrugalComputing

The model takes into account the trends in embodied carbon (in fact, that is the main contribution of the research) and in global carbon intensity of electricity generation. For the latter I am assume we stay on trend, although it is conceivable that the energy hunger of AI would slow down the decarbonisation of energy production.

#FrugalComputing
(5/n=5)

I have also looked at the IEA scenarios. They use a bounded growth model, and their projections are only until 2035. But even with that model, in the worst case we'd see emissions of more that 1 GtCO2e/y (about 10% of the global carbon budget) purely from AI.

#FrugalComputing
(4/n)

The graph shows an extension of the McKinsey high-growth scenario until 2045. This is exponential growth, 10x in 10 years, 100x in 20 years.

Note that Michael Dell said we'd need 100x in 10 years, and Altman said something similar about growth in AI servers.

So if the expectations of the AI industry would be realised, by 20245 AI alone would cause more emissions than the entire global carbon budget.

#FrugalComputing
(3/n)

The aim of the paper is to model the emissions ensuing from the growth scenarios put forward by the IEA, McKinsey and the AI industry. We created a very detailed model for this. Here is a sneak preview of one of the graphs in the paper.

#FrugalComputing

(2/n)