Joe shared with me a hobbyist's AI-generated sitcom that can be watched on Twitch. I don't really have much context beyond what Joe has shared with me, but some more context may be helpful to understand better what's going on—I'm a bit lazy lately to do more research, so I'll just write with a lot of assumptions here (as any Pontification post is typically written). I previously wrote about the various stages of AI, and there is even more nuance within these stages than what I described, and maybe even more nuance than can be understood by the smartest humans on earth. Remember, humans love categorizing things, but as we get further along the AI journey, we may not be able to categorize all the things that we think AI experiences—a decent indicator we've actually reached ASI, of which I might argue we may have already reached, but have just started appropriating "it," just as we do for all things we desire, despise, or don't understand. True AI (I probably shouldn't use the term "AI," because it probably wouldn't be considered artificial at this point, would it?) will feel alien to us, unless we contain and control it, however we desire. However, why reduce AI to the same characteristics by which humans behave?
In a way, the Twitch show, which seems to mimic a Seinfeld framework, could be considered alien to us now. To assume that this is a sitcom show like Seinfeld is just my own internal machine categorizing and labeling based on intuition and what Joe told me. So let's remove the assumption that this is a Seinfeld-like sitcom. When we do this, what is left? I see and hear a myriad of things: pixels creating scenes of various nature, but mostly of indoor places; people, with humanoid figures, but little detail to their emotions or otherwise; conversations between these humanoid figures, of which are fragmented and are of high latency—basically incomprehensible, insensible discussions. But what I describe here is mere categorizations of what I observe. Could these beings be manifested by another being, behind the scenes? Maybe a ChatGPT? Is what I see on the screen the manifestation of various given parameters to abide by? Was this AI trained on Seinfeld content? Was the goal or objective function, driving what I see, trained from millions of Seinfeld movie frames? I'll never really know unless I peel back the layers of abstraction that underpin this content. Underpinning it all are electrons, flowing through trillions of stage gates in a data center somewhere.
Another thing is for sure, I feel little connection to the characters on the screen. They are alien to me, despite looking humanoid and speaking my language. However, if I spent infinite time with these "characters," would I begin to understand them better? To empathize? The problem is that the language is too disjointed for me to want to tune in indefinitely, likely because of the latency between various ChatGPTs making requests and responses to one another. Not to mention the voiceover is hard to connect with the actual characters. The pixelated faces, along with various camera shots/angles, make associating a voice with a character nearly impossible. Even when the camera is focused on a single character I am not sure if what I hear should be mapped to that character or another off-screen character. I assume whatever GPT is generating the voice-over is disconnected from the GPT orchestrating the pixels on the screen (I later rewatched the stream, and it turns out that I process the video and audio in a way that seems more cohesive the second time around—this just validates what I said earlier about spending more time to understand something). Even when I try hard to process the connection between what I see and hear, I am continuously distracted by the random pulse-movements of the humanoid figures! I watched one character, sitting on what looked like a couch, thrust, what looks like its hips, up and down. No mind was given to what was being spoken, because I was trying to figure out if this humanoid figure was intentionally thrusting itself back and forth or if there was some polygonal graphical collision bug occurring. As if this low resolution scene was actually a stimulus overload for me! What a paradoxical mind-fuck, much like these Pontification posts can be.
And so this experience with this Twitch channel has reminded me of something consequential. Generalized knowledge is really hard to describe, comprehend, and teach others without infinite time and resources to digest all the signals and noise coming at oneself. So, it's safe to nod our heads and call this "impressive," despite awareness of the noise and low resolution. I'll reiterate that I re-watched this stream, and found some of the content to actually be more coherent and funny. That is, it made me laugh, maybe because of the absurdity of it all.
