As technology continues to advance, so do the challenges and ethical dilemmas that come with it. One such challenge is the growing concern around the use of Artificial Intelligence (AI) in academic plagiarism. With the increasing availability of AI-powered tools and platforms, it has become easier than ever for students to plagiarise content and evade detection. This not only undermines the integrity of academic institutions but also poses a significant threat to the education system as a whole. In this article, we will explore the dangers of AI in academic plagiarism and the measures that institutions and individuals can take to prevent it.
Ok, confession time – the paragraph above was written by Chat GPT, after I entered the prompt ‘please write me an introduction to an article on the dangers of plagiarism in academia in relation to AI’. It’s ok; it lacks a hook or engagement point, it seems a bit unsure of its target audience, but it reads fluently and makes a surprising amount of sense. For anyone unfamiliar with Chat GPT, it is an AI chatbot which generates text based on a prompt. GPT will produce anything from a short conversation to a Harvard-referenced essay on the development of technological determinism in late twentieth century ambient music.
Fortunately, the essay Chat GPT generated in response to that last prompt was requested by me out of curiosity, rather than submitted by one of my students. I am happy to report that the resulting text was rather circular, repeated the title a lot, made up a few facts (Vangelis and Eno collaboration?) and seemed to avoid producing any actual content. Poor GPT also got very confused by the prompt to Harvard reference the resulting text, and talked about the famous University quite a lot. I’m glad that Chat GPT is still a way off from doing my students’ homework for them!
Fighting plagiarism
AI has quietly become the latest concern in the eternal fight against plagiarism, or AI-giarism as it has been dubbed. Chat GPT has recently introduced a watermarking scheme for their AI-generated content. The idea is that any text generated by large language model AI will have a fairly predictable pattern of word choices and syntax. Much like Turnitin or similar plagiarism checkers, this feature could be used to indicate the origin of any potentially AI-generated text.
In the music world, AI has been making quiet waves for some time. The first AI-generated songs have been recorded and released, with a similar level of success to my essay experiment. For anyone unfamiliar with Chase Holfelder’s AI generated ‘Rudolph, the All-Gracious King’, please treat yourself to a slightly disturbing dystopian view of a Christmas hit.
The current AI offering
Where AI has been making more serious, and more successful, inroads in music is in the sphere of production. iZotope has recently released Audiolens – an AI-powered desktop app which provides a reference analysis of tracks from any audio source including, for the first time, streaming platforms. Audiolens is designed to sit next to Ozone 10 or Neutron 4, iZotope’s mighty AI mastering and mixing apps. Audiolens collects data on key sonic characteristics of your selected reference track, which can then be imported into Neutron or iZotope. The engines then match the elements of the track you are working on and adjust them to sound like your reference.
While Ozone has been a quiet secret in the back pocket of mastering engineers for some time, Neutron is establishing itself as a go-to tool for mix engineers. Once a musical ensemble has been captured in a multitrack recording, the first job of a mix engineer is to create a tonal balance, ensuring that each element of the mix can be heard, and that the individual tracks sit together in a convincing way. Typically, the tools we reach for are EQ and compression to achieve this tonal balance. The next stage is the creative part of the mixing process; for a classical or jazz ensemble, the job usually ends at making the artists sound the way they did in the room, with no exaggerations and as few reductions of dynamic range or spatial qualities as possible. For other genres, the journey begins here.
User experience
Consider the creative decisions behind the use of pitch correction software in a Travis Scott track, or the creative use of stereo in Radiohead’s Glass Eyes, and you get some idea of the blank page staring back at the mix engineer. Neutron is a suite of high-quality mixing and mastering tools which can be used in a traditional way. The real power in this suite comes from the Mix Assistant; once booted up, this machine learning algorithm will make both technical and creative suggestions on your mix which can be based either on a specific reference track or, even more impressively, on genre or style.
In contrast to the amazing labyrinth of machine learning algorithms underpinning both Neutron and Ozone, the user interface is as simple as you could imagine. A series of on-screen prompts guide you through the process of balancing and creating your ideal mix, asking you whether you want to reference a track, take a genre suggestion, what platform your final mix is intended for, and how impactful you want the final result to be.
Prior to the release of Audiolens, those of us who spend any amount of time on the mix/master-side of the process would habitually hoard a collection of reference .wav files on our computers. While the debate around audio quality in different file types will continue to rage, I can think of several examples where my students have noticed elements of a mix which are not audible in lower quality files. This fundamentally changes the way we mix for streaming platforms; if a particular sonic quality is important, it has become the job of the mix engineer to understand how different platforms, different headphones or speakers, or different formats will translate your mix. Check out how different the same recording can sound on Spotify and YouTube, and you’ll see what I mean.
Audiolens allows the mixer to reference tracks directly from streaming services, helping me as a mixer to understand how the glorious high-quality recording I’m working on will be translated by the process of file compression, Bluetooth compression, and speaker intermodulation which it will go through before reaching my listeners’ ears. The tools at my fingertips to make this translation are a Sculptor, Compressor, Dynamic Equaliser, Exciter, Gate, Trash (a distortion filter) and Transient Shaper. The unmask mode, rather like a driving assistant in a high-end car, points out and corrects the technical flaws in my mix, revealing points where different instruments don’t sit together in the ideal way. The library of instrument types allows some user input to the combination of dynamic EQ, compression and panning which Neutron uses to enhance the final sonic result.
Suitable for the classroom?
With much of the Level 3 syllabus for Music Technology based around a series of mixing tasks, the availability of this suite poses an interesting ethical question. Like any AI, Neutron can be a powerful tool to speed up a mix process or get an engineer out of a tricky corner. I don’t know a single working musician who would rely on Neutron to produce a final mix, and the manufacturers did not intend the platform to be used in this way. However, the potential for this to form at least part of a student’s coursework portfolio is significant. Unlike GPT’s watermarks, elements of a mix are expected to conform to predictable patterns of sonic choices and musical syntax. It is much less easy to spot the involvement of an AI mixing.
At around £200 for the full mixing suite, Neutron 4 is unlikely to become a staple of the classroom any time soon, although pared down elements versions are available from £32, with subscription versions of the software from Splice, and several other sites.
There’s no doubt that AI is going to become part of our technical and creative tool kits as working musicians. It already has. In addition to Neutron, AI-powered amp and mic simulators are becoming staples of recording studios, lyric and chord generators are becoming staples of writing studios, and I even recently attended a gig live-streamed by an AI. Likewise, we stand at a crossroads as educators: AI could become the next issue in the ongoing battle against plagiarism. Alternatively, we could look at creative use of AI in our own teaching practice. Several universities offer tuition on Ozone as part of their mastering programmes, with the expectation that the AI is combined with other tools in the final production. One creative writing course recently suggested students use Chat GPT to create jump-off points for their own writing. We have the choice to embrace these tools as part of our pedagogy, or risk becoming victims of them.
