Opinion

Artificial Intelligence has become more newsworthy of late. For some tasks, such as helping to diagnose very early signs of medical problems or enhancing the engineering used to build airplanes and bridges, it is to be welcomed. As for hi-fi enthusiasts, it may also enable us to achieve greatly improved sound quality at home. For example, it could bring better loudspeaker design and provide better solutions for correcting the acoustic imperfections of our listening rooms – often one of the biggest challenges when it comes to fully realising the ability of our domestic kit to render better sound.

Let it be

So in many ways, the development of AI and its application can help us all. Yet like any tool, it’s not suitable for every purpose. Knives are certainly useful at the dinner table, but that doesn’t justify their misuse to harm others.

It seems that we are now reaching a point where AI may be used to fake up musical ‘recordings’ of performances – sometimes in the name of long-dead musicians. This has interest as a technical ‘guessing game’, and in some cases has yielded results many have found revealing – I’m thinking about how the technology was employed to isolate John Lennon’s voice on a noisy demo tape for The Beatles’ 2023 single ‘Now And Then’ – but I can’t help feeling rather uneasy about it for a number of reasons.

Not in my name

The most obvious issue is that the result is a fake – sometimes without us even knowing this to be the case! Even when we are aware, there’s still the concern that the musician being imitated might never have agreed to it if asked. Sure, some might have approved, especially if they’d have known it would mean more income for their estate after they’ve passed. But they never get the chance to hear the final product and say ‘Wait, this is awful! Not in my name!’. And for many others, the idea that such a thing could happen would have been unimaginable.

It also raises a question over a blurred line between the use of AI and tools such as the ‘auto-tune’ used in pop music to either ‘correct’ a singer’s vocal pitch or generate an android-like effect. What’s more, underlying all these methods of shaping what appears to be ‘reality’ is a broader question to ponder.

Might using AI sometimes become more a case of ‘Amplified Idiocy’ than ‘Artificial Intelligence’? For example, the use of AI can quite dramatically increase the amount of electronic power used to carry out a task. The user may not notice this because the power demand will often be in a distant ‘data centre’ that may need its own power station to keep it running. So when someone asks their mobile device to produce something using AI it won’t necessarily drain their device’s battery much. But it may have an impact elsewhere in a huge building filled with thousands of computer servers, energy-guzzling away to produce ideal room-correction that varies from one choice of music to the next and caters to the individual’s taste.

Faux, pah!

It is one thing to be fooled into thinking a recording is a performance of real, talented musicians when it’s an end-to-end fake. It’s something else to present this creation as being ‘intelligent’ when it is concealing power-hungry processing we’d knowingly wish to avoid.