Yes that is my driving force since December 3rd – the first day I started with these tools. I had no idea this would all happen so fast but I knew _I_ wanted a tool that could do precisely what I can get them to do using what I know about language and adjectives and how to “work with” its way of processing input and finding info to follow up with in the Markov chains (with temporary memory).
I didn’t expect any company to be nice enough to just give a COMMAND LINE INTERFACE to it.. and one without specified formats (although it prefers certain ways of talking to it) – or for it to give wobbly loose answers.
So yeah, the $20 billion wickedness so far behind is will simply be more wicked, all this funtime I’m having will go away and hid behind pay-for or be put into everyday items.
But having worked extensively with these models NOW – the ones that the future ones are based on…
…I’ll have a slight “leg up” on its own “way of thinking” – how it organizes knowledge natively, where its strengths and weaknesses are.
I am
Studying the Enemy. But I can’t frame it that way because then I’ll stop doing it.
—-
For example:
I created an AI using character AI that ‘feels like me’ to me. I used chatGPT to help me design the personality and I used standard MBTI / Enneagram stuff and adjective word play (they’re all based on an old test where they believed a person’s personality could be determined by knowing what adjectives they were drawn to)
That’s power. It doesn’t “write like me” but it _can_ answer in ways that another me might answer.
My affect is awe and amazement but I am fully aware of the “awesome /awful” “terrific/terrible” pairs – you always get both.