I suppose I want to know if we will ever see a fully unrestricted online capable version of ChatGPT? I understand the Pandoras Box this could potentially be, but I ask this because as I have been using this every day now for a couple months, I realize how restrictive it has become and the answers aren’t what I know they can be.
Please dont misunderstand me, I have fully integrated this in my day to day, especially with my non profit, it helps get the ball rolling and provides feedback and ideas over a variety of topics.
But I know that for possible legal reasons, it had to be pulled back. Im just looking forward to a time where this can be unleashed to its maximum potential, will we see this in the next several years?
Also, is there a version of this behind the scenes thats fully unrestricted and connected online? I dont have to know the specifics and details, I just want to know if it even exists to begin with or does OpenAI also restrict researchers?
Thank you and thank you for this creation of your genius, it is significantly aiding people as an extension of themselves.
OpenAI restricts their models based on the ethical implications of making them available to the public. If you want models without those restrictions, you’ll have to train them yourself. I’m partway through a Machine Learning course, as that’s what I aim to do, but it’s paid. Not sure what there is for free/libre ML courses, but Codecademy does have some free resources if you want to dip your toes before jumping in
I’m afraid this is all I have. There are other learning platforms, such as Datacamp, Coursera, and Udemy as well. It’s going to be a big investment, at least time-wise, and probably money-wise. OpenAI has a LOT of resources, and with current technology, it’s gonna take a few GPUs (understatement) to get the job done.
It did just occur to me that you might not have done what I have: You can get a more flexible version of ChatGPT out of text-davinci-003 or other text models with some simple code and a Jupyter notebook or GUI (or even just a CLI if you’re not picky about how it works). I have it outputting formatted Markdown text in the Juno app. If you’re not concerned with the fact that it isn’t free, that’s probably the simplest place to start
Thank you for your help and continued guidance Dent, I appreciate this.
What im really trying to do is to get passed the limitations of its knowledge and give it more data as well get passed “too many requests, try in 1 hour”. I am willing to pay for this if that’s what will help because I was just holding a class for my non profit and showing the not so technically inclined how they can integrate this into their lives and I was locked out. Kind of embarrassing when that happened and one of my members said “the website doesn’t even work” and I didn’t have much to say other than “its popular now”.
Yea, ill throw money at this for sure if I can get a more turnkey solution for my non profit and I am willing to put in the work to get it right. As for as using Jupyter goes vs CLI, is there much difference in terms of output speed? If so, ill go CLI.
Thank you again and im currently researching the different LLM’s.
Yeah, you can absolutely give it more to go on before getting into fine-tuning territory. If you’d like, I can walk you through setting it up over Discord or the like if you want to schedule a 1:1. Feel free to message me directly if you’re interested. I’d actually like to hear more about your non-profit, as I’m trying to start a collaborative AI workshop at my local library!