Economically, AI has been compared with the dot com bubble - only more negative. As a figure of speech, cost of RAM goes through the roof and one consequence of that, sales of new PCs and phones, through the floor. Apparently even Microsoft noticed that as the security updates for Windows 10 have been extended for another year.
While some may be impressed with AI's ability to ransack the web and make useful summaries, colleagues at DIY audio forum asked it to design an amplifier. The outcome can only be described as total rubbish, AI confabulated even symbols nowhere existing in electronics engineering.
One of Murphy's laws applies to AI: Everything that can go wrong will go wrong, just much earlier than expected.
True, and it goes for everything: "good enough for even the most sensitive (ears) or critics" would result in a much lower resource drain - including energy. But that's what most environmentalists completely ignore.
Several years ago I posted a simulation of a very simple amp, sub-ppm THD even at 20 KHz. A member didn't believe that and made it - upon hearing the result, his wife with "golden ears" said "you can keep it here (in the living)". Cost of the amp: below $100, can be wired in less than a day.
The best class D amps when run at low power produce in the 0.00X THD and for "golden ears" still produce too many unwanted artifacts at higher volume. What I will construct is a class A where dissipation at low volume (desk speakers) is below any class D or AB. Calculation showed it will be a net energy saver. Less than 20 parts / channel (easy maintenance) and single digit ppm THD at 20k. Good enough for the rest of my life and that of those inheriting it.
FET is only better than BJT when the driving impedance is high: when not, Darlington or Skizlai is much better (20 dB in THD and several volts in output swing).
Opamps provide such a low output impedance and TI produced several, able to perform at ~40V - enough for 20W maximum power at 8R. Operating point can be run via series R of output inductor so one can set dissipation as required via supply voltage control. Trick would work with tubes just as well but you can't overcome the power for the filaments. But a semiconductor version can run at dissipation as low as 3W and produce 0.6 W audio at ppm THD (desk amp use). These days, getting the HiFi inductors is THE issue: I ordered mine (custom made version) in Poland. When leaving politics, could be interesting: https://ogonowski.eu
"Our best minds can’t catch the complexity of the problems we are facing"
Even if they could, outcomes would come in probabilities. I know that he next lottery numbers are an unordered 6-element subset of {1,..x}. Yet, it does not help me.
OK Ugo, I'll be a mitochondrion, or, rather a Chloroplast, which is pretty darn similar, but performs photosynthesis. This is a non-trivial difference, of course.
Another smart guy who does not comprehend thermoeconomics? "I'm shocked, shocked..."
Economically, AI has been compared with the dot com bubble - only more negative. As a figure of speech, cost of RAM goes through the roof and one consequence of that, sales of new PCs and phones, through the floor. Apparently even Microsoft noticed that as the security updates for Windows 10 have been extended for another year.
While some may be impressed with AI's ability to ransack the web and make useful summaries, colleagues at DIY audio forum asked it to design an amplifier. The outcome can only be described as total rubbish, AI confabulated even symbols nowhere existing in electronics engineering.
One of Murphy's laws applies to AI: Everything that can go wrong will go wrong, just much earlier than expected.
Good, functional, elegant circuit-design takes actual intelligence.
No understanding can be too deep...
True, and it goes for everything: "good enough for even the most sensitive (ears) or critics" would result in a much lower resource drain - including energy. But that's what most environmentalists completely ignore.
Several years ago I posted a simulation of a very simple amp, sub-ppm THD even at 20 KHz. A member didn't believe that and made it - upon hearing the result, his wife with "golden ears" said "you can keep it here (in the living)". Cost of the amp: below $100, can be wired in less than a day.
A "sand amp", I presume.
Here is the latest Chi-Fi at that price, made from a Texas Instruments chip
https://www.ebay.com/itm/297488869631?itmmeta=01KW59ZC2TDGW44MVHVGP8TXQ5&hash=item4543b7e8ff:g:ucUAAOSwQKBlgmIX
The best class D amps when run at low power produce in the 0.00X THD and for "golden ears" still produce too many unwanted artifacts at higher volume. What I will construct is a class A where dissipation at low volume (desk speakers) is below any class D or AB. Calculation showed it will be a net energy saver. Less than 20 parts / channel (easy maintenance) and single digit ppm THD at 20k. Good enough for the rest of my life and that of those inheriting it.
FETs?
FET is only better than BJT when the driving impedance is high: when not, Darlington or Skizlai is much better (20 dB in THD and several volts in output swing).
Opamps provide such a low output impedance and TI produced several, able to perform at ~40V - enough for 20W maximum power at 8R. Operating point can be run via series R of output inductor so one can set dissipation as required via supply voltage control. Trick would work with tubes just as well but you can't overcome the power for the filaments. But a semiconductor version can run at dissipation as low as 3W and produce 0.6 W audio at ppm THD (desk amp use). These days, getting the HiFi inductors is THE issue: I ordered mine (custom made version) in Poland. When leaving politics, could be interesting: https://ogonowski.eu
Wow. I and Emad Mostaque learned two things:
1. The Landauer limit
2. Fuck with anyone as smart as Ugo Bardi
"Our best minds can’t catch the complexity of the problems we are facing"
Even if they could, outcomes would come in probabilities. I know that he next lottery numbers are an unordered 6-element subset of {1,..x}. Yet, it does not help me.
OK Ugo, I'll be a mitochondrion, or, rather a Chloroplast, which is pretty darn similar, but performs photosynthesis. This is a non-trivial difference, of course.
Another smart guy who does not comprehend thermoeconomics? "I'm shocked, shocked..."
;-/
Hipótesis: "Las obras más grandiosas siempre quedan a medio construir cuando una civilización cae"
El moai más grande jamás tallado aún está en la cantera.
Es una hipótesis que requiere verificación pero no es improbable por lógica.