Due to the slow motion collapse of Twitter, we have all been slowly losing touch with various people and sources of information. So, of course, I am losing convenient contact with various people as they no longer appear in my Twitter feed. Sure, I hear things – people tell me when there are people making veiled jokes about my criticism at sessions, and I see when my blog posts are just flat out plagiarized out there (not like any of them are that original anyways). I know no one on any “Future of AI” session would have the guts to put me (or any other deeply skeptical critic) on their panels to respond.
There also seems to be a loss of other things as well, like what some people in our field call “getting Downesed.” I missed this response to one of my older blog posts critiquing out of control AI hype. And I am a bit confused by it as well.
I’m not sure where the thought that my point was to “deny deny deny” came from, especially since I was just pushing back against some extreme tech-bro hype (some that was claiming that commercial art would be dead in a year – ie November 2023. Guess who was right?). In fact I actually said:
So, is this a cool development that will become a fun tool for many of us to play around with in the future? Sure. Will people use this in their work? Possibly. Will it disrupt artists across the board? Unlikely. There might be a few places where really generic artwork is the norm and the people that were paid very little to crank them out will be paid very little to input prompts.
So yeah, I did say this will probably have an impact. My main point was that AI is limited to not being able to transcend its input in a way that humans would call “creative” in a human since. Downes asks:
But why would anyone suppose AI is limited to being strictly derivative of existing styles? Techniques such as transfer learning allow an AI to combine input from any number of sources, as illustrated, just as a human does.
Well, I don’t suppose it – that is what I have read from the engineers creating it (at least, the ones that don’t have all of their grant funding or seed money connected to making AGI a reality). Also, combining any number of sources like humans is not the same as transcending those sources, You can still be derivative when combining sources after all (something I will touch on in a second).
I’m confused by the link to transfer learning, because as the link provided says, transfer learning is a process where “developers can download a pretrained, open-source deep learning model and finetune it for their own purpose.” That finetuning process would produce better results, but only because the developers would choose what to re-train various layers of the existing model on. In other words, any creativity that comes from a transfer learning process would be from the choices of humans, not the AI itself.
The best way I can think of to explain this is with music. AI’s track record with music creation is very hit and miss. In some cases where the music was digitally-created and minimalistic in the first place – like electronica, minimalist soundtracks, and ambient music – AI can do a decent job of creating something. But many fans of those genres will often point out that AI generated music in these genres still seems to lack a human touch. But on the other end of the musical spectrum – black metal – AI fails miserably. Fans of the genre routinely mock all AI attempts to create any form of extreme metal for that matter.
When you start talking about combining multiple sources (or genres in music), my Gen-X brain goes straight to rap metal.
If you have been around long enough, you probably remember the combination of rap and metal starting with with either Run-DMC’s cover/collaboration of Aerosmith’s “Walk This Way,” or maybe Beastie Boys’ “Fight for Your Right to Party” if you missed that. Rap/metal combinations existed before that – “Walk This Way” wasn’t even Run-DMC’s first time to rap over screaming guitars (See “Rock Box” for example). But this was the point that many people first noticed it.
Others began combining rap with metal – Public Enemy, Sir Mix-a-Lot, MC Lyte, and others were known to use samples of heavy songs, or bring in guitarists to play on tracks. The collaboration went both ways – Vernon Reid of Living Colour played on Public Enemy songs and Chuck D and Flavor Flav appeared on Living Colour tracks. Ministry, Faith No More, and many other heavy bands had rap vocals or interludes. Most of these were one-off collaborations. Anthrax was one of the first bands to just say “we are doing rap metal ourselves” with “I’m the Man.” Many people thought it was a pure joke, but in interviews it became clear that Anthrax had deep respect for acts like Public Enemy. They just like to goof off sometimes (something they also did with thrash tracks as well).
(“I’m the man” has a line that says “they say rap and metal can never mix…” – which shows that this music combination was not always a popular choice.)
One unfortunate footnote to this whole era was a band called Big Mouth, who probably released the first true rap/metal album by a rap/metal band on a major label (1988’s Quite Not Right on Atlantic Records). As you can see by their lead single “Big Mouth” it was… really bad. Just cliche and offensive on many levels. FYI – that is future Savatage and Trans-Siberian Orchestra guitarist Chris Caffery on guitar.
Big Mouth is what you would have gotten if our current level of AI existed in 1986 and someone asked it to create a new band that mixed rap and metal before Run-DMC or Beastie Boys had a chance to kick things off. Nothing new or original – and quite honestly a generic mix of the two sources. Sure, a form of creativity, but one you could refer to as combinational creativity.
I don’t say this as a critique of AI per se, or as an attempt to Deny Deny Deny. People that work on the AI itself will tell you that it can combine and predict most likely outcomes in creative ways – but it can’t come up with anything new or truly original in a transformative sense.
On to mid-1991, where you probably had what was the pinnacle of rap/rock collaborations when Anthrax wanted to cover Public Enemy’s “Bring the Noise” and turned it into a collaboration. The resulting “Bring the Noise 91” is a chaotic blend of rap and thrash that just blew people’s minds back in the day. Sure, guitarist Scott Ian raps a little too much in the second part of the song for some people, but that was also a really unexpected choice.
There is no way that current AI technology could come close to coming up with something like “Bring The Noise 91.” IF you used transfer learning on the AI that theoretically created Big Mouth… you might could get there. But that would have to be with Anthrax and Public Enemy both sitting there and training the AI what to do. Essentially, the creativity would all come from the two bands choosing what to transfer. And then you would only have model that can spit out a million copies of “Bring the Noise 91.” The chaos and weirdness would be explorational creativity… but it all would have mostly come from humans. That is how transfer learning works.
Soon after this, you had bands like Body Count and Rage Against the Machine that made rap metal a legitimate genre. You may not like either band – I tend to like their albums just FYI. So maybe I see some transformational creativity there due to bias. But with a band like Rage Against the Machine, whether you love them or hate them, you can at least hear an element of creativity to their music that elevates them above their influences and combined parts. Sure, you can hear the influences – but you can also hear something else.
(Or if you can’t stand Rage, then you can look at, say, nerdcore bands like Optimus Rhyme that brought creativity to their combined influences. Or many other rap/metal offshoot genres.)
So can AI be creative? That often depends on how you define creativity and who gets to rate it. Something that is new and novel to me might be cliche and derivative to someone else with more knowledge of a genre. There is an aspect of creativity that is in the eye of the beholder based on what they have beholden before. There are, however, various ways to classify creativity that can be more helpful to understand true creativity. My blog post in question was looking at the possibility that AI has achieved transformational creativity – which is what most people and companies pay creative types to do.
Sure, rap metal quickly became cliche – that happens to many genres. But when I talk about AI being derivative of existing styles, that is because that is what it is programmed to do. For now, at least, AI can be combinationally creative to spit out a Big Mouth on it’s own. It can come up with an Anthrax/Public Enemy collaboration through explorational creativity if both Anthrax and Public Enemy are putting all of the creativity in. It is still far away from giving us Rage Against the Machine level transformational creativity that can jump start new genres (if it ever even gets there).
Featured image photo by Yvette de Wit on Unsplash
Matt is currently an Instructional Designer II at Orbis Education and a Part-Time Instructor at the University of Texas Rio Grande Valley. Previously he worked as a Learning Innovation Researcher with the UT Arlington LINK Research Lab. His work focuses on learning theory, Heutagogy, and learner agency. Matt holds a Ph.D. in Learning Technologies from the University of North Texas, a Master of Education in Educational Technology from UT Brownsville, and a Bachelors of Science in Education from Baylor University. His research interests include instructional design, learning pathways, sociocultural theory, heutagogy, virtual reality, and open networked learning. He has a background in instructional design and teaching at both the secondary and university levels and has been an active blogger and conference presenter. He also enjoys networking and collaborative efforts involving faculty, students, administration, and anyone involved in the education process.