You are viewing a single comment's thread from:

RE: LeoThread 2024-11-04 12:58

in LeoFinance13 hours ago

Oh so there are people who disagree with that?

I wonder why they would, cause the AI system uses pattern recognition to give out fact. The more context you give it during training, the better it can link up stuff and produce a result you'd find more meaningful.

Yes without proper context it may still produce a somewhat good answer, but to give indepth relevant answers would depend on the indepth context we feed it.

Anyways taskmaster did mention something to me yesterday about that which I would link to you in a second.

Sort:  

yeah.. most people are very afraid of change. Denial is probably a keyword

I think we're talking about two different sides of the concept of context here. I'm referring to providing context in the prompting of AI, while you're referring to the training of it. It is my understanding that when it comes to training models, the amount of data and quality of data is more important than the context the data is structured in - that is sort of what the training process does by default.

Then you can add in additional specialized data on top, which is what LeoAI will be utilizing. In the case of LeoAI, all data that is being added to threads (and hive in general) is continuously being fed into a vector database that the pre-trained model LeoAI is using, can draw data from in its responses and actions, in addition to the general base training the model has already gone through (using all sorts of data from all over the internet).

I guess I had a gab in understanding there. Thanks for clarifying this for me.

Really appreciate it.