Performing learning in the AI age

Some of the commentary around the potential impact of generative artificial intelligence on learning runs something along the lines of a lament: that we are producing a generation of young people who won’t actually be able to really do anything for themselves.

From there the AI fans tend to spin off into fantasises about how this won’t matter as technology will take on more of the effortful drudgery, freeing up time for the moments when human insight really matters.

Conversely, the doomers ask whether the nature of learning is such that the ability to do basic things builds into these high-level mastery capabilities, and in removing the chance to learn and practice these building blocks we negate the very possibility of this mastery manifesting itself.

What’s clear to me is that people who hold these positions haven’t been to an open mic night recently.

One, two… one, two

I live in a small town in Gloucestershire. Off the top of my head I can think of six regular open mic nights happening within fifteen minutes of my street. This last couple of years has seen an explosion of opportunities for musicians and performers to take a few minutes to present something they have been working on to an audience – while not always an enthusiastic or appreciative audience, at least one that is present. Sometimes you get a free drink too.

From the guy who got a guitar for Christmas and has just learned a Green Day song to semi-professional musicians trying out challenging new tunes they have written or learned, the only real entry requirement is that you are brave enough to try entertaining a pub full of people with nothing but your musicianship, stage presence, and that microphone (occasionally with the support of the house band).

So why would anyone put themselves through this on a wet Wednesday night in a pub just off the A46? Fame, adulation, monetary reward – these are just a few of the benefits conspicuous by their absence. For those I’ve spoken to (this includes me, though I promise you these findings are not solely based on autoethnography), the reasons they do it are threefold:

• To do something they enjoy: in this case, making and sharing music
• To challenge themselves: to practice something that they may not yet be confident with
• To network: to meet other like-minded individuals who have the same interests and goals as themselves

But those aren’t just reasons to dust off the old dreadnought and look for some chord charts – they’re reasons to do pretty much anything that takes us time, effort, and an openness to personal vulnerability.

How Suno is now

If you’ve not played with any of the numerous AI powered song generators out there, rest assured that they are loads of fun. A few words from you can shape a professional-sounding piece of music with a decent amount of radio-friendly polish. Anything that gets people into the idea that music is something that you can personally shape the creation of is fine by me – tools like Suno are a gateway into understanding more about what you like and dislike in music.

As such, I would venture that the majority of the music tools like these generate is never shared, and seldom played more than once. This is a playground – a place to experiment with ideas – but it is neither a replacement for or a threat to musicianship as practiced in the traditional manner.

Whereas many enjoy using generative AI tools, they neither present a personal challenge to be overcome, nor offer a networking opportunity. They present something that feels like a finished product (or a step on a way to a finished product). And that’s it. If you enjoy doing something – if it is a part of what you value as a person or something you wish to do in the future – you actually do the thing. And you are generally very interested in how the thing is done rather than the end product. Intrinsic, rather than extrinsic, motivation if you will.

Tools like Suno are a pathway into thinking about the creation of music – but they don’t really offer a bridge across into developing your own musicianship. The prompt window can distinguish styles reasonably well, but doesn’t respond well to directions about chord inventions, rhythmic accents, or choices of scale as a basis for melodic invention. And why should it? You don’t need that knowledge to use the tool (just, I add cynically, a subscription.)

Do you even lift?

Law professors will take pride in telling you that most of their students do not become lawyers – law is not a vocational degree. However, pretty much all law students have an interest in law and justice, and whether or not their end goal is to practice law their course will concern “practising” law in the broader sense. Students on any course are committing to do particular actions multiple times in order to gain mastery – be that cannulating a patient, calculating the results of a chemical reaction, using evidence to build a convincing argument, or taking a solo over Coltrane changes.

The only reason to sign up for a course like this is if you enjoy (or think you will enjoy) the thing itself. Otherwise you either spend three years doing something you don’t enjoy just to spend forty years doing something you don’t enjoy – or you just spend three years doing something you don’t enjoy. And who would choose that, even if they slept each evening clutching graduate employment statistics and LEO data?

Part of the concern around students using AI tools to write essays is that it is “cheating”. But even if an undergraduate cheated their way through an entire three year degree their only reward (and I use the word loosely in today’s job market) is the chance to start a career doing stuff like the activity that they didn’t enjoy enough to do themselves at university. With a side order of an imposter syndrome above and beyond that experienced by most young people.

I’ve never once seen anyone turn up to an open mic night with a song they’ve written using AI. Part of this is cultural (backing tracks are generally a “no” – you are looking for karaoke for that) but a part of it is that the sheer effort involved in preparing something to perform only really makes sense if you really care about the thing you are performing. And of course if it is the kind of thing you enjoy doing.

A busy night, so only one song

I’ve been thinking about the reason that there are so many open mics in my small town – and why they are always so busy. Somewhere in our often justified concerns about student use of AI we have lost sight of the fact that doing something once isn’t really the issue: it is the ability to do it again on demand, no matter the situation or the circumstance. There’s a downside to our current glut too: there are a dwindling number of occasions and venues offering paid work for musicians in the traditional way, so for many musicians the open mic is the only realistic chance to play in public.

If you decide – as a student, or as a person – to devote a fair chunk of your life to a particular practice or body of knowledge then we can take that as an indicator that you want to do the work. And if you become proficient, you will want to demonstrate that – and you will want to be around others who are on the same journey, for inspiration or encouragement or even a bit of healthy competition (if you haven’t nailed those twiddly bits, rest assured someone will note this and offer to show you how).

In more traditional academic terms this is the beginnings of a community of practice and a confidence about your place within it, but it is not too fanciful to build a bridge over to ideas of belonging.

The price of joining the community, the price of proficiency, is diligent practice: that you can do what you say you can do and you can prove it. AI might be a tool that can help you practice – or help you understand what you want to do – but it isn’t going to get you through that first song alone in front of that microphone. But once you have done it for yourself it is likely that you will want to do it again, and do it better.

But don’t do Wonderwall. Please.