Raw LLM Responses

Except LLM's aren't a physical thing. They aren't a screen. They are ephemeral math equations that emerged from widespread GPU's and Memory. They only just emerged from the cocoon. It's more apt to compare it to a cell phone in the early 80s, that still has 30 years of maturing for it before it gets to those diminishing returns. We have no clue what AI would look like in 10 or 20 years, and even if it stagnated completely and only got faster/cheaper, that alone would be enough to push the field forward massively, because the #1 bottleneck is token throughput right now. Using AI is like just sitting and staring at a progress bar, which reminds me, I gotta go check on my prompt.