Gen AI is a control wasteland. Cue a Jon Spencer Blues Explosion to shred the slop.
The year was 2003, maybe 2002 or 2004? I was driving Jamie Lidell, then of Super Collider fame, to a gig taking place on a boat. This was before “Little Bit of Feel Good” made him a household name.1 He crooned through Paris traffic but for what on earth did he need the insane amount of luggage we hauled through a packed Gare du Nord? On location, the Seine rolled—we spilled insane gear. FX boxes, mini-synths, tweaked pedals, echo chambers, and a Gordian knot of cables spilled, daisy-chaining the mess. Then came mastery. Lidell transmogrified into the breathing live sequencer of his own music, ripping and rebuilding his sound—tearing, stitching, and tweaking his voice, his drums, his keyboards. That was music and raw control. Wizardry, plain and simple. Hence the smorgasbord of knobs, pads, and faders—back when controllers weren’t king. A raw jolt for anyone craving soul in I/O’s slop.
"Synthesizer". The name says it all. Promising automation, creation on tap, much like Gen does for AI. A myth, turned misconception. Machines do it all. Vintage synth demos pushed presets. They flopped—Klaus Schulze aside.2 808s and MPCs birthed techno and hip-hop, genres totally unforeseen by their makers. Emergence ruled. The real shift was not automation but interface. New controls calling for new mastery. Will this happen again? Synth to AI, the parallels are striking. Engineers ultimately oblivious as to the real creative use of their tech. So what if we pushed past the ruling chatbox—what’s hiding beyond?
The secret lies in interface not automation. New controls allow for new ideas. Parameters, tokens, weights, and benchmarks? Total red herrings. They’re as useless for creativity as arguing over memory or graphics in video games. Forget the noise. Ask real questions. What can you pull off with creative control, mad skills, and room for happy accidents? How do you turn this tech into art? What can you shape hands-on? How much space is there to move and tweak? Can it feel as free and fun as spinning tracks?
Shocker, the answer lies in Websim's unbridled chaos, not ChatGPT's clean little box. Think of your tech as a sandbox not an enclosure. Allow for the unexpected. Let the users play. They are the ones who land the breakthroughs. Much like community tinkering elevated Ableton. An update is mandatory. A design shift from the sterile to the playful. So far, all GenAI interfaces have shared the same flaw. Their textboxes completely deny user control. It’s an engineering fault—unease with creativity. Cluelessness, or just fear, about the whole shebang. To these engineers, creativity is a mere variance allowed in the output, nothing more. How do you solve this? Strife or play. One means overhauling controls with tech, by doing so, asking for empathy they never had. The other is better: ask, implement or hack manual override. Or just a monitor. Why not go all King Tubby on it and witness the results of experimenting with what we understand or not.3 Give us knobs, buttons and a feedback loop. Ditch the Chatbox and we will hit it out the park.
To nail that home run, we need designers to rethink the interface. We need them to solve for what they ignored: creative control. How? By asking engineers to let go.
Community tools show the way. Image generation has its very own synthesizer: ComfyUI. A daisy chain just like Lidell’s; hooking image generation to control the effects. It’s quirky tech; going mainstream means getting playful. The first company to make it the MPC of image gen will win big. In short, a Comfybox. Interoperable, just like MIDI. A sandbox that plugs into any model.
In the end, a real instrument.