SydneyMusic does not use generative AI or AI platforms. We are an AI-free organisation.
This policy applies unilaterally to all functions of how we work, and includes:
Specifically, we do not use generative LLM-based technologies that rely on scaled tech platforms built from stolen human work. We do not think that any efficiencies that they might create makes up for the longer term downsides which we take very seriously.
We also don’t platform certain uses of generative AI – for example, music with obviously AI cover art is not added to the weekly gig guide playlist.
When we launched this project, our tagline was “No ads, no algorithms, no shit!”, promoting an alternative to relying on social media feeds for music discovery. At that time (early 2022) ChatGPT was still 8 months away from launching and AI was still a comparatively niche discussion topic.
Today we believe as passionately as ever that a growing reliance on algorithmic presentation of information has been a net negative for humans socially and culturally. Facebook, Instagram and TikTok’s design philosophies have had a notably damaging effect on local community ecosystems.
We value the time and energy that humans spend researching, learning, creating and doing — and that includes the process of discovering music and gigs. Algorithms that make decisions on how information is processed and sourced without transparency or control remove human agency.
AI now extends the conversation about algorithms beyond “content consumption” but the same core principles apply: when you take shortcuts in work, you remove the hundreds of tiny decisions we make every day that refine our rationale and ideas, and help us to improve, and invite curiosity.
We take pride in the gig guide, which is the byproduct of hours of human work every week. The time spent researching, learning and thinking about our local ecosystem is shown in the quality of what we produce for you — and we do that so you can take control of your discovery and have a complete view of what’s happening in our amazing city.
Put simply: it would make the gig guide worse, and our testing bears this out.
We are confident that the guide’s secret sauce lies in the many hours we’ve spent poring over information about artists, venues, promoters and events. The human judgement involved in the research we do has consistently surpassed the accuracy and capabilities of LLMs. AI chatbots and LLMs authoritatively produce responses that seem right, unless you already know enough about the subject to tell when it’s wrong.
Most of the hyper-local info that makes our gig guide useful to you doesn’t actually exist anywhere else on the internet. If you ask ChatGPT what’s on in Sydney tonight, it will pull information from, among a few others, our gig guide – info researched and curated by humans – and use it to create a sort of guide-slop that sounds right but is not actually useful info. We have consistently proven that for the granularity and specificity of the information presented in our guide, it’s mostly wrong, and it’s usually wrong about the big basic stuff too. (If you don’t believe it, try it right now – ChatGPT informed us a real band that was actually in town last week is playing tonight, at a venue that does exist but which they did not play at once during their entire tour. Looks about right. Is 100% wrong.)
Everything that your AI bot of choice spits out is based on information that was the result of work done by a human. Even big platforms like Songkick and BandsInTown are a mix of automatically “scraped” info and human input, and they can’t always tell the difference between, say, a burlesque night and a gig, or between a covers band and a local act performing their own songs. Understanding the distinctions between these things matter: companies that prioritise efficiency over precision are going to miss the details. That’s a major element of what’s being lost from our culture as organisations and businesses “pivot” to prioritising AI output and processes, largely in order to get rid of those pesky, expensive human workers.
The information our human team members are researching exists on a huge range of different platforms, websites and social pages — and sometimes not on the Internet at all — all laid out differently, all displaying the basic info we need for the guide. Even if we managed to write a script or build a tool to collect all that information, humans would still need to double-check it to make sure that this bit is actually the venue name and not the name of the band night, or work out whether the headliner is listed first or last. That’s the stuff that makes our gig guide good and useful, and AI can’t use context clues to work this stuff out. Humans can. And if we need to have that step in the workflow anyway, then the automation isn’t actually adding anything. Once you factor in building, maintaining and fact-checking an automated tool, all that extra work usually negates the benefits it is supposed to deliver.
We’re nerds and we love tech. We don’t reject technical evolution just for the sake of it.
We’re particularly concerned that the adoption of generative AI is focused on the products built by a limited number of scaled (and mostly US-based) tech companies. We don’t like that these platforms are a bit of a blunt instrument: they’re trained by gobbling up a huge amount of human output, and then the responses don’t really have any method or transparency in how their responses are produced — they’re largely a black box.
There may be positive applications of Large Language Model (LLM) technology, but we don’t like the principles that currently steer its evolution via corporate investment.
If we were to make use of LLMs in the future, we would like to see the tech being used in ways that celebrate and promote small-scale design principles that favour individual ownership rather than centralisation of resources.
For example, we would like to see tech solutions that:
It seems that just about the entire AI hype cycle is centred on only the above. We believe that smaller, non-centralised applications of this technology allow for agency and control to be returned to the operators and users of these systems.
Until we start to see trends move in this direction, we’ve decided to sit the movement out entirely. We don’t think we stand to lose anything by making this decision.
We welcome feedback. Contact us