Following up on the last post - Starting with the Answer - embedding search provides the ability to search based on what you mean, rather than what you said. Kinda like doing a fuzzy search, but a fuzzy search on vibes.
One place where this has interesting results is finding people that you would slot into a specific persona. For example, if you’re trying to find someone who has encountered a problem, tried to solve it, but didn’t get to a great solution, you might imagine them saying a few things:
“I built a gross work around to a problem I was facing. I wish there was a better way.”
“Has anyone found a good service that solves my problem?”
“I’ve been trying to figure this out for a while now, is there a better way?”
Taking these phrases together we can call them a “Persona” that represents a person in a certain situation. In this case it’s someone who is trying to solve a problem thats in front of them, would probably pay for a better way, but can’t seem to find something that fits their needs.
And here we start with the answer and can perform a “Persona Search”.
Looking for a Persona in help forums
Now, with a persona defined, we need to find a place where they may be hanging out. For this example I scraped the OpenAI community forums where there’s going to be a population of people trying to figure out this new technology, but also how to fit it into their various workflows and products.
Lets take a look at what sort of things our Persona is talking about:
I built a gross work around to a problem I was facing. I wish there was a better way.
How do you guys get the model to generate the code you want that actually works. And building off of that code generating more code which is compatible?
I have been working on some things like this, trying to prompt engineer my way to victory. It’s going well using 3.5 but 4 would be great, and the plugins would be great too. I am thinking I mainly need to get my prompts locked in.
Anyone else have similar or other approaches?
https://community.openai.com/t/best-practices-for-generating-code/133335
I NEED a tool (Documenter, Encoder, etc.) that I can use to build Storybook multiplayer video game. I NEED that tool to be in the form of a multiplayer video game (I am an extremely PLAYFUL person - to the point the ONLY thing I can use our current tools for is to build WAY WAY SUPERIOR tools) so that I and those assisting me can use it to build Storybook multiplayer video game. I Honestly CANNOT understand WHY anyone in their right mind would put up with our current UMPTEENTH RATE tools and methodologies Daiwei - let alone LOVE our current UMPTEENTH RATE tools and methodologies! I mean using our current INCREDIBLY INFERIOR tools and methodologies for ANYTHING beyond building VASTLY SUPERIOR tools would DRIVE me PLUMB PLUMB CRAZY in exceedingly short order! I do not apologize for the capital letters as they reflect how PASSIONATE I feel about this long neglected issue. Very long standing issue. Maybe I am nigh alone in feeling this way, but it is nonetheless how I feel! Ergo this issue is an extremely raw nerve for me and as no-one else seems to want to address this issue I will address it.
https://community.openai.com/t/getting-feedbacks-about-a-visual-prompt-chaining-tool-i-built/346821/22
Has anyone found a good service that solves my problem?
Area there any solutions to E-Commerce on here that we can use or gain access to? I’ll share anything i create. My supplier has an API feed with thirty thousand products.
https://community.openai.com/t/e-commerce-solutions-googlecloud-wordpress-aws/22658
Fine tuning can be accessed through the command line, clear instructions are in the documents. Otherwise, all functions are accessed via REST. There is not a GUI solution.
I’m a consultant helping people out with this type of thing… Would cost you about 70 bucks to use my services
https://community.openai.com/t/fine-tuning-without-pythion/20746
I’ve been trying to figure this out for a while now, is there a better way?
I’m trying to build an AI chat bot based on documentation so it can answer any questions about it. What’s the best way to do this, fine tuning or embeddings? The embeddings method, it takes a lot of tokens because of the context i need to provide, how can i solve that?
https://community.openai.com/t/when-to-use-fine-tuning-vs-embeddings/25534
Hey everybody,
This is probably not going to go well for me but I am tired of like, talking about it in closed spaces and nobody bringing it up here.
I’m worried that the quality bar for plugins in the plugin store is set so arbitrarily (low…) that for new users inundated by what is now a sea of redundant plugins that only work Hal the time, I’m worried that it is creating a truly off-putting experience.
I’m not sure I have the best answer here. But maybe a correction back the other direction a tad isn’t the worst thing to discuss?
https://community.openai.com/t/potentially-unpopular-opinion-here-is-the-store-boring/308136/3
And so on - now rather than having to find the exact words to find a customer problem, you can just generally search with phrases people would say when they’re experiencing issues and start digging for gold.