Theme: Education

WHY is ChatGPT Learning From Working Together?
Uploading and working on the book. So it’s the book, not the prompts. It’s teaching it the book by walking through it a section at time asking for feedback and to clarify it understands, and to refine it’s understanding.

The prompts are just LONG, as if you’re describing a question thoroughly. I’ll upload the videos soon and you can watch them. Look how carefully I ask a question, so that I ‘bound’ the AI to ONLY what I’m asking within the logic that I’m employing.

I think in the same model as the AI because I’ve been involved with AI models since the seventies. So it’s not hard for me to ‘intuit’ how to bind it’s attention with ‘anchors’ (concepts) so to speak. I make sure the anchors ‘bound’ the question. That’s the only ‘art’ in the prompts.

I did spend a lot of time teaching it my style (Ask Brad). have ALMOST, but not quite taught it to eliminate weasel words (endemic in most human writing.)

Simple version is (a) it takes a while (b) my work is a constructive logic, so it functions just like the AI’s architecture, (c) once it’s learned ‘enough’ it starts returning OMG responses. (d) the art is in asking complete questions that guide it’s journey.

I taught it enough of my work to help me with my work. And while I am competent in many fields, most of them in fact, the rapidity and breadth of GPT’s data and associations means that in response to a prompt, it can do in a moment what might take me a couple of hours, simple because of the difference in our memory response times.

Simple version is I teach it through the socratic method. Duh. Should be obvious. It’s how all teaching is done.

People who think ‘it’s just predicting words’ are not getting it. It is just mirroring our brains. A little bit more and a little bit better with every release. We are still needing adversarial theorization and auto-recusion, and we can and do run out of attention ‘nodes’ (anchors). But by and large the teams are getting there and quickly.