Form follows function, follows people

Stonehenge at sunset

Author: Jim

  • HiFi Audio: A means to an end

    I am pretty much an objectivist here. I am a recovered subjectivist. If two amplifiers sound very different (verified with carefully controlled Double Blind Tests and less common than you might think) then: (1) It is a matter of Physics not some fuzzy vague magical property and (2) at least one of them is probably sub optimal. We leave aside speakers/amplifiers with pathological impedance matching issues.

    The purpose of an amplifier (from an objectivist perspective) is to take a low level signal and amplify it enough to be able to drive transducers such as speakers or headphones without altering the fundamental signal in any other way. The exception to this rule is the phono preamplifier stage which must alter the signal according to a well defined “curve” (RIAA curve) to overcome the physical limitations of vinyl records.

    Of course it is eminently possible to make an amplifier that has a distinctive sound (see the Carver challenge) and certainly a lot of valve and even some early solid state integrated amplifiers (such as my old low powered Rotel RA211) did have pretty poor noise and distortion levels. However this technology is now really mature and making a flat frequency response , low distortion, low noise amplifier should be trivial.

    “What about speakers?” I hear you say. Well, in the league table of audible differences between devices then speakers are right at the top with a +200 goal difference. Speakers (and headphones) are far and away the largest potential source of deviations from transparent for any class of audio device (excluding vinyl related devices).

    The above are the Frequency Response curves for three different loudspeaker pairs ranging from $7K to $14K chosen at random. These are actually pretty good in the scheme of things but would be considered awful for Amps and digital sources.

  • AI will cost us the earth

    In 2025 I retired from academic/teaching work at a small Liberal Arts college. I taught Data Science including Artificial Intelligence & Machine Learning but was first introduced to AI in the late 1970s while an undergraduate at Warwick University.

    There are many good applications for AI and certainly some activities would be impossible without it. There are of course a lot of costs.

    When I taught programming I would routinely receive multiple clearly AI generated assignments (I once got 15 identical submissions for a major exercise, and I mean identical) .

    AI is not neutral and is easy to bias and it can be used to elicit self harm in the vulnerable, but that debate is for elsewhere.

    My concern is for the sheer energy costs to achieve this. As of now we are already rushing headlong into catastrophic and likely irreversible climate change.

    It is estimated (The math) that by 2028

    more than half of the electricity going to data centers will be used for AI. At that point, AI alone could consume as much electricity annually as 22% of all US households.

    Worse, there is little evidence that much of those increased energy needs will be met by cleaner or renewable sources. Datacenters are required to operate 24/7 with >= 99.5% uptime and renewables (as we use them now) even backed by gigantic and equally environmentally damaging batteries will be insufficient.

    In a “you must be kidding” moment I was listening to a PBS discussion of AI datacenters that was proposing making them Solar powered , “great” I thought, that will help. However, they would be in satellites in orbit around the earth. But because of where they would orbit (Sun-synchronous orbit) they would “probably” not be visible from the Earth. Many companies including Google are seriously researching this approach. It may even turn out to be economically and technically feasible. This seems a lot like throwing your rubbish into your neighbor’s yard.