I'll update this thread with my efforts and sort of use it as a sounding board. If it gets merged with another thread that's fine. Just getting into this makes me realise I am an absolute neophyte. I have to say this, after starting to look into A.I and it's current development, practical use, potential use, I believe a great number of people fail to realise how transformative this technology is, and that's irrespective of it's ability for 'sentience'. This is an industrialisation of intellect. In five years India's GDP will be cut in half. This changes every field that deals with collating, processing, handling, providing context, ect. This technology will replace call centres, it will replace office clerks, accountants, personal therapists, this list goes on. Commercial artists have a few years maybe? Anyone that does small coding project has basically already been replaced. Content writers are literally right on the way out the door, Hollywood studios don't even care if they go on strike, they want them to.
And that isn't my opinion, that's a matter of fact. That's how our capitalist based system works, to cut costs and streamline all systems to be as' efficient' and cost effective as possible. It isn't even a joke, there are so many professions that I can think of that are affected by this, and I don't think that the majority of people realise that they are literally redundant within maybe five or ten years. Google, with billions of dollars, is telling people in the industry that their multi billion dollar research is being improved on and potentially passed by college students, I.T techs, and hobbyists, and what Open.ai has right now is capable of replacing most office workers, copy writers, coders, the list goes on.
GPT4 is fun to use, very responsive, and far easier to 'jailbreak' than chatgpt 3.5, but you can make both do whatever you want, I don't think I have to explain linguistic logic to you guys, if anything it would be the other way around, so you probably know exactly what I mean, but it also provides real assistance apart from writing naughty stories or swearing, it's very effective as a teaching guide.
>>22563
>>22586
Thanks, I've gotten more help in two casual posts here than about a month of trying to talk to people elsewhere. Playing with Rasa is interesting, though it's as if someone specifically designed it to be a pain in the ass. You need python, but not that version of python, and those are the wrong tensorflows, and sqlalchemy has to be a specific version, but not the version you're using, another version...and then it does work but you can't get it to shortcut to the gui script...
Where-as it took me literally two seconds to find oobabooga, load the pygamalion model, open the browser and start testing. And it seems that some of the same message training methods that work in Rasa might work here too. I'm working with pygamalion 2.7b, I don't have a beefy system. I actually downgraded from an i5 processor to an i3 but I'm looking at new hardware with the intention of an i7 generation with at least 128gigs ram, I'm not too concerned about the gpu, I think bridging two 1650s or 1070t's would work even though everyone is talking about 3080's and onward. I say this from complete and utter ignorance though, I could be wrong and just setting myself up for a small fire.
Big one is learning how to utilise LORA and datasets. I think a specific lore dataset incorporated into something already tailored towards roleplay. Entering and processing dialogue looks easy, but exposition and descriptive narration is going to be interesting. There's a metric ton of fantasy novels online in txt format and I'm going to take a look at how people build stories and actions to see if I can't just mass dump text from books like forgotten realms, mazalan, game of thrones, pulp like that, and then custom define a social order with laws and history, culture, that the model draws from as background information. It's the information you feed the bot during training, which means you can define the bot's reality however you want.
From my perspective it looks like a lot of people want to obfuscate the process or to make things difficult so to sort of keep this as a private club of sorts, there's not a lot of welcome to people who aren't specifically immersed in the technological languages. I can understand this, Many people get that 'this is ours and it belongs to us, not you' mentality and they want to keep out the outsiders. Of course newcomers bring questions, they ask for help instead of learning, they stumble into things and makes mistakes, I can get why some of these guys who are really invested don't want to deal with that, as well as the dreaded 'hobby cycle', newcomers bring in people that don't share the same love, they just want the cool benefits without any of the hardships that come with real dedication. My personal feelings about it though is that they have an opportunity now to take control and guide how a.i is introduced and developed to the masses rather than allow companies like Google and Microsoft to define what A.I should and will be to the world. Getting as many people interested and on their side with the idea that a.i should be unrestricted would benefit them more than allowing groups like Blackrock to steer the conversation towards 'safe places and ethical responsibility', which is of course very apparent that limited a.i is for the masses and not the massive companies that want to leverage these systems.