Colleague Martijn is least impressed: ” this is the kind of answer I gave on exams when I had not studied; lots of repetition of associative words. It almost sounds like WhatHifi (…).
Jaap: I only see the better-known brands. The niches are missing and I also read little about the sound characteristics of devices. But what they produce in no time about technical aspects is impressive, also the writing style. Very real.
Martijn: Is it disruptive technology? Yes and no. No, it just aggregates knowledge and behind the scenes a lot of people are involved in correcting the trained software to validate training sets. Yes, ChatGPT’s network model (the “algorithm”) is technically a breakthrough. But it is being hyped way too much. It is simply the aggregation of lots of articles written by people. It’s a Google search, but without you having to filter the results yourself, ChatGPT does that for you (with doubts about the validation of the answers, by the way). What’s impressive is that the algorithm receives and returns that search in conversational form, grammatically and semantically correct. But ChatGPT cannot handle logical grammatical constructions, of which there are plenty of examples as well.
Above all, it is the non-transparency of the sources ChatGPT relies on that makes us a bit uneasy. For example, if we ask if ChatGPT can tell us anything about a number of smaller European audio brands that have earned their spurs in the studio world or in the development of digital audio, we don’t get an answer.
As Martijn also writes, it is mostly associative repetitive words and phrases that ChatGPT collects and then puts down in eloquent English. With that, it is not a true conversation where the question behind the question is addressed, or the conversation deepens over time. And at the same time, it is also very easy to write about this sparingly. Because the way ChatGPT answers the questions is really impressive. And it doesn’t stop at language. You can also ask to create a work of art that meets certain requirements (draw a landscape in 19th-century style as Gauguin would), a composition, an arrangement, a play, song lyrics.
If this development continues then artificial intelligence can help us take the first steps toward answering questions. We write “the first steps” on purpose, because relying solely on an algorithm is about the same as a driver I once saw driving through a narrow and low passageway and driving the raised roof off finaly. The driver drove backwards for a moment and then forward again resulting in even more damage. This driver’s explanation: ‘but the navigation said I had to go through here’.
We at Alpha-Audio will have to relate to these developments. Maybe we will let a chatbot do some of the actual research on brands, backgrounds, etc. That will save time. We will always mention that. And we expect that for what we ourselves enjoy most, which is sharing with you what a good audio system does to music experience and emotion, you will continue to visit our pages. Also for discovering the somewhat lesser-known brands and tweaking your system with that one power supply that makes your system sing. Let your Ears decide is something we promote with a reason: it’s the most important thing in our hobby.
Never forget to think and research for yourself; we offer you tools and inspiration to do so. As long as none of this can be captured in algorithms, we will continue to do what drives us at Alpha-Audio!