[JoeMark]

I haven’t been able to get the Llama_2_70B to run.  I think those 70B LLM’s require a serious GPU like an A100.  Let me know if you have it working?

With that, I am on a similar quest to see who is applying/testing/comparing outputs of SuperPrompts to the open source LLM models.  I fear we are all investing time and energy into ChatGPT to only have the rug pulled out from us.

It makes sense to me that focusing on the uncensored models might be the best place to start.  As we are learning, OpenAI’s guardrails appear to be lowering the performance of the model… I assume that is true for any model.

[JoeMark]

@markerubin I have a MAC M1 too. It’s not suitable for running local Ai.

With that, is there a youtube channel or location online that tests various hardware configurations (preferably laptops)?

[JoeMark]

Very interesting approach. I think we need a follow up motif  structure. For example, templates for customer personas, demographics, psychographics, TAM, market driven vs product driven strategies, unique selling propositions and emotional selling propositions, pricing structure including potential competitive analysis etc while maintaining brand voice across all those results.

From there, messaging strategies for priming your target audience beachhead market (I would look at several source references like the book Pre-suasion to develop an editorial & social media calendar with specific messages based on target personas and buying behaviors)

you get the idea. there’s much more to this model as we integrate existing marketing analytics from other data sources like CRM, ecommerce, web & social media metrics

I’ll give this some thought and post some ideas.

[JoeMark]

Roxxanne – not spending a lot of time prompting. While fascinating, it strikes me that prompts have a shelf life. My attention is toward integrated workflows with Langchain, Hugging Face, Pinecone, etc. Starting to see ChatGPT is more of a platform than a solution itself.  The same for GPT4ALL.

[JoeMark]

Thank’s Brian

You already share a lot. No pressure no obligation. Just curious as thought I also read something about an impending article on FinGPT and other very interesting topics. Wondering if I was missing something?

 

FinGPT was very buggy when I installed it. Ran out of time trying to fix it. I may try again tomorrow.

[JoeMark]

Thanks for all the help. I was having a problem with the downloads failing. Several attempts go them working.

[JoeMark]

I have downloaded the additional models but they didn’t load into the application.  Anyone know what I’m missing? Local files doesn’t seem to be working right either.

Running GPT4All on a Mac.

[JoeMark]

interesting. So while Jordan is no longer practicing psychiatry, this could essentially open a virtual practice. What could go wrong?

[JoeMark]

Thanks Larry – I like how you broke out each of the 5 traits as personas.

I am thinking of using all of the 5 traits within in one persona however applying various weighting factors to each. Just like normal people

likewise I will be creating multiple personas each with their own Big 5 personalities… perhaps other dimensions like age, gender, culture, etc

I’m curious to see how they collaborate when I set them on a complex task.

[Author]

Posts