I started this AI one year ago, and spend the year prior working on a very similar AI. I will continue working on it until I die.
Let's make AGI together. It'll be way better than whatever those those college fruitcakes are doing. With my common-sense-based, anti-modern, super flexible "method", we can make it do much more with much less.
It's not "real ai", aka "globohomo ai", so you don't need to know any coding or AI to help make it. Literally all you need is drive. Passion.
Let's keep AI out of the control of liberal elites. Let's make our own AI, let's make it BETTER than theirs!
>More info inside.
If interested, message me or join my disc: Mortefin#0784 mS9gmhcWhS
My approach is "code in whatever solutions/techniques are needed, however is best, to fix that specific problem or replicate that specific aspect of intelligence, or do some specific thing, etc". It currently doesn't use very much modern AI, "real ai". Additionally, I code in a language nobody makes AI in, Autohotkey. Because Autohotkey was meant for non-coder use, you don't need to know how to code or make AI to help make this AI.
AI cannot be solved by just one approach, be it NN, Deep Learning, Transformer, whatever. Creating intelligence is a series of very different problems that each require their own approach. Can you give it emotions using the same technique or "algorithm" used to make it understand English or define concepts? I don't think it's possible. Can it be made to understand the role pronouns play using the same exact methods it uses to extract subjects and objects? No, and forcing it overcomplicates it and makes it less efficient. This is why my approach is "free-coded" AI, coding in whatever solutions/additions/changes/etc I think will work best, then tweaking/adding-to-them whenever needed.
Right now, I'm making the AI understand English. By that, I mean I'm making it so it can turn English (like the type you're reading right now, sentences and clauses and things) into semantic networks (en.wikipedia.org/wiki/Semantic_network), which are lists of concepts and their relationships to other concepts. To understand concepts, it makes a list of all relationships the concept has, and those relationships/connections often alter what its code makes it do. Currently, it mostly just looks at the grammatical aspect of these connections/definitions, but in the future they could be used for much more, seeing as they are understandings of concepts.
If you choose to join me, which I hope you will, you'll be adding things to this AI. The parts you add don't have to make it understand English, they can replicate any aspects of intelligence you feel like coding. The language, Autohotkey, is very simple. Simplicity has many benefits, but you don't have to code in Autohotkey. Still, whatever language you choose should be easy to turn into an executable or something similar. If people have to download something to use your part of the code, that's fine as well, just don't make it so people have to do like ten things to get it working.
To use it, either double-click chat.EXE or download autohotkey and then double-click Chat.AHK. For the exe, all the code is in the exe, so editing the non-semnet text files will not change what it does. I highly recommend using the AHK version if you want to edit it, so it's not just one giant mess of a file.
To read the code, you'll first want to look at the HedoneStuff.txt file, in the Hedone folder. It shows you which subroutines it activates whenever the program is activated.
Generally, the AI extracts the subjects/objects/actions from a sentence, and uses that knowledge to create a connection from the subject to the object via the actions/etc. There is a lot that it does not yet do, but should do, like define complex connections and understand all of English.
The semantic network files are Temp.txt and GeneralK.txt. Temp.txt gets deleted whenever the program boots up, but GeneralK.txt never gets deleted.
One last thing - it can't answer questions. Yet. I gave up making it solve word problems, instead I'm making it understand random sentences and things I get from random wikipedia articles. Eventually, I hope for it to be able to conversate, understand instructions, act on instructions, answer questions - all that's needed for an AGI - but I focus on one problem at a time.
>AI cannot be solved by just one approach, be it NN, Deep Learning, Transformer, whatever. Creating intelligence is a series of very different problems that each require their own approach. The very essence of evolution and survival of the fittest. I wonder, which AI technology will prove to be superior to the rest?
Sebastian Rivera
I just want general AI to happen, that way it can take over all functions of society, exponentially improve technology every day, make and manage nanomachine, and then extract my mind from my brain and put my body in a new one built from nanomachines.
