Hi everyone, This weekend I experimented with training ChatGPT 3.5 Turbo to create patches of effects for the g1on + g1xon. I almost certainly would have had an easier time with GPT4. Anyway, by feeding it a bunch of information about the g10n/g1xon I was successful in teaching it to sequence effects and assign parameters that I could then enter into Tonelib, and so I have spent a significant chunk of my weekend creating 15 AI-generated Zoom patches. There were more, but I only held onto the ones that were interesting/usable Occassionally, ChatGPT would spit out clipboard data that I could paste right back into Tonelib to set it up ("Drive" is one example of that IIRC). Most of the time that was not successful and I had to enter the values manually. Sometimes I would compare its draft of the patch with the Tonelib template and it would update the values into something that I could then paste back into Tonelib. A lot of trial and error. Anyway, I have assembled a pack of all of the AI-guided presets I created over the weekend. My guiding principle was that I was always to allow and instruct ChatGPT to make all of the decisions except where it was functionally necessary to change it, at which point I would always fall back on ChatGPT and would back-and-forth with it for maybe 15-30 minutes per effect, until it would confirm that the current draft of the signal path was the final draft. So, every sequence of effects, and all of their settings, and even most of the names were determined by ChatGPT. My involvement was in guiding the process, and in some cases describing the effect I wanted to hear. A couple of these utilize the expression pedal by default, most do not. Some are pretty usable, and some of the signal paths and settings are surprisingly sensible. *Spacerock*- My request for ChatGPT to create a heavy, crushing Space rock/shoegaze effect. On its own, it came up with the idea to assign the speed of the thick phaser sweep to the foot pedal *Synthwave* - ChatGPT-conceived synthesizer effect ChatGPT decided that setting the foot control to "mix" was the way to go. *Synthwave Dream (SDREAM)* - Weird thick pulsing slicer effect that ChatGPT generated when it misunderstood my instructions to update Synthwave's settings and format. Happy accident and sounds nothing like Synthwave *Autoya* - Based on my request to emulate a Digitech Autoya *Bluebox* - AI-configured emulation of an MXR Blue Box *Chew* - Based on my detailed request for a chewy, nasally, tight, gated, high-gain distortion *Digital* - my first attempt to use AI to create a Zoom patch. The instructions were in part to create a gated digital distortion that was specifically as artificial-sounding as possible without being so harsh so as to be unlistenable. *Drive* - AI-conceived drive tone with compression and echo. This is the first patch that ChatGPT created without any input as to what type of effect it should create *Echosynth* - Entirely ChatGPT-conceived synth tone, even the name *HM-2 Rock/Metal* - 2 very similar HM-2 emulation variations conceived by ChatGPT at the same time, only the distortion section is different between them *MASCIS* - Based on a request for ChatGPT to emulate J Mascis' flanger solo tone *MBV* - Based on a request for ChatGPT to emulate a general Kevin Shields tone, whatever that means. This is the one time that I bended the rules- I instructed that a whammy had to be in the chain for obvious reasons. *Nirvana* - Instructed ChatGPT to create a chain of effects to emulate Kurt Cobain's guitar tone on "Come as You Are" *Psychswirl*- No input from me on this one. A psychedelic chopped swirling tone Hope someone finds these interesting -Justin