I'm sorry folks, as always, the previous research burst, which would have been great if I hadn't found a job, just stalled. My work is very physical and involves a lot of moving around, so my organic research downgraded to after-work hobby tier and thus effectively stalled for over six months.
Now, instead, because I don't have the mental energy to read but I do have the strength to type, I'm constructing a human-multi-agent development tool that will use currently existing software architectures, applications and functionalities as agentic plugins for a research tool-creation center that will improve the data 'warehouse' management, storage and organization with as of yet non-existing but desperately wanted functionalities and customizability created with the back-end philosophy of this whole blog, which is already being used as its training material as an ideal example of one person's writing style, range and interconnections of interests, ways of keeping exact and preferred uniform bibliographies, etc.
With my little "geet", you become your own agentic research team that eventually will start producing unthinkable novelties for a researcher willing merely to write up everything that pops into their mind connected with research, even random stuff that makes no sense, just write about own interests, theories, hypotheses, random thoughts, hunches, feelings, plans, etc. until there is a sizeable corpus (e.g. 100,000 words or 600,000 keystrokes) and just let LLM-s simultaneously train on and start looking for research to train further until you have a virtual research-you with whom to talk about specific stuff further, have them talk to your mock agentic characters that you can completely make up or base on sources that are rare and start building their case and personality yourself by talking with them about specific quotes, using other quotes, etc. Anything is possible!
Not only will it eventually learn your research idiolect, it'll talk to you as any other AI agent on the market currently would, but instead of generalizing from public information it would instead specialize for your personal information that you feed it yourself: you teach it first and then it teaches you. You don't give the business machine man a few tokens, have it find out what's the most mediocre answer to your question and be done with that single use; you write up your whole life if you want to, your most deeply burning research questions, everything you *want to know exactly but can't yet* and you give my magic abstract semiotician program over a part of your soul, and it will create you *the exact research application you actually need* because *you* design it my writing up what really need. My little geetie-pooh will grow into a monster that'll know what's in every library, every archive, every collection willing to hook up for a bit.
My "AI" does not pretend to be intelligent or conscious: it is a hive of python3 plugins in a trenchcoat trying to help with the most useful thing in the world: research.
It'll be great! I'll post a summary 100, 1,000, 10,000 or however many pages long human-readable text explaining all of my ideas with as much necessary context that you're willing to donate for me to generate it in a push of a button. It'll be a tight 100 or 300 pages at first and selective because I haven't even begun the NDA process (need to set up OÜ first). So, very interesting times we're living. The world's first actual semiotics factory, with online participation and local offices to temp semioticions from all around the world may actually be a realizable future.
Idea has been brewing for 12 years, let's see if it's safe to drink. If I die, it wasn't a suicide - I have a good-paying job and a life partner 'bout to wife up. I'm not the internet's own boy. I'm a boy who reads old books and shows you the middle finger from atop a boulder while smiling knowingly.
0 comments:
Post a Comment