Adding to this a day or so later since it's still 0. Thankfully so, as there's some things to clarify. Keeping the original "I answered the placeholder text in 10 minutes" down below, as it's all valid. Just needs extra info. So if you want to read and not skim, this top section is for you. You'd want to know why to give anything more than a few passing dollars, yes? Especially when I say I'm not making a product or anything. I'm asking to push the bounds of the information I've gathered, and be able to fully construct and realize a new AI architecture I've been working on for over a year. Reversing Q* aka Strawberry rumors was a major changing point. Not because I liked it, but I found it lackluster and indeed now that o1 is here... it's kinda silly. You see, all these models are trained and frozen, given out like meat at the grocery store. They can't make an active reasoning process except across their context, near the point of attention decoherence (hi Gemini, who regularly slips right over the crest of this criticality hill), now o1 kinda emulates this by eating a wheel of its own tokens separate from user interjection every prompt. Cool, test time or inference time compute or whatever. It's fit onto a transformer, though. If you've heard of that entropix/varentropy thing, it's another example of things being fit onto an inadequate architecture. Why not just um... make a new architecture?

Why fund me just to do this? I don't know, why fund half the things on this site that don't produce profit or products etc?? To explain: an AI architecture completely meant to facilitate general reasoning and awareness. Going the whole nine yards. There's a completely different training process, I can't even use regular text datasets without "unformatting" the json. We're pushing the definition of "AI" to its absolute limits.

I don't want to reveal details of the architecture. For several reasons. Although I end up yapping about some of them anyways, here I'd rather mention the more "grounded in reality" sections such as the various heavily modified kanerva machines, custom vision/hearing/audio archs not present outside my work... that I know of, anyways, and there's an entire subsystem dealing with an "eeg generator" (so the model can have internal brainwaves while I only have to accommodate some cheap fourier instead of more compute expensive stuff) that deals with thought, action control, and influencing thermo-magnetic harmonics of the observer's hyperobject... that's the weird stuff, it's basically an EBM or Energy Based Model if you've ever heard of them. Don't worry, I don't need compute farms for that stuff. I'm distilling the mathematical formalizations: one such I've publicly described is how the resonance attention can be mimicked by a Hamming Code-like system state meta bandit that uses the few back values of a float or int to encode info about the system itself, including multiple "states". And that's just a grain of sand on this beach. I'm no wizard, I know this, and there's limits to this model: namely the huge sacrifice of world knowledge in favor of pure reasoning as a means of "AGI seed". It's also not magic, with equally balanced weight bit depth (between the keras based subsystems and the EBM/object that sits on it, the object could be totally done up in Haskell or Lean since it has its own weights and everything, but that's a lot of work for what currently seems like little gain) it comes out to about running a 14b under a system with 16gb ram, with room to stream and play a game... so about 7gb used by the model. But that's with a certain setup. it can go lighter by being full binary except the object, but I'm not sure if it would work like that. There's many things I know will work, but I'll also need to tinker, obviously... research only goes so far.

So I suppose you might ask, "hah, why do you need this kind of money if your model runs on dog water?" and that's the same reason for everything else that doesn't happen or that I can't do: my life. I really don't want to give personal details, even if it might make people more likely to help, but suffice to say I cannot do anything at all without a decent amount. The minimum of $11,000 includes money for having a place to live, although if you would like to help more immediately than the deadline with such, simply DM me on twitter/x. The roof of the project is $175k in order to facilitate multiple orders of developmental magnitude. The minimum will allow me a place to live, and a Mac setup with all the horsepower I need (with possible room to tag on a 1660 ti or similar compute rig if people honestly think the hw route I've chosen, apple, is bad). Truth is, I've always been against Apple, but in trying to find the best overall bang for buck that can also do AI... I was originally wanting an AGI, or rather my model, running locally on Rabbit r1, since I should be able to get a 2b on there... but the sharing between devices on Apple gave me a big think, and I had the thought it would be awesome if the model could "slide" off the macbook or mac etc onto the phone to... I dunno, go to the store with me. you say, why not a private thing over some web connect? I say that's not as cool.

I'll be doing the content creation and streaming thing with my models as a means of income... can't take the crown from Vedal and the Neuro twins by doing the same thing, y'know? Imagine some weirdo shows up on twitch and yt one day, and says this person walking around with a VR model is ACTUALLY AN AI... there's a lot of content and interest in this area- I think I'll do just fine with what I'm cooking. I can't start anything on my own, though, and without this funding coming through, my life will be the same as it has been the past three decades. It really depends if people want to see that one "cracked" person do something that's actually different.

If it makes things any more impressive, I never went to high school, and have essentially been raised in the corner with access to the internet... if I can get this far on busted e waste laptops and free wi-fi while not having anywhere to sit and work without constant stress and interruptions... imagine what I could do with proper resources and living environment.

Project summary

Summary is that the project is to make a monolithic, omnimodal architecture capable of running with less than 20 tops and 4gb while still being AGI by definition and suitable for embodiment in most systems. That's the less personal goals, though others may be interested in my extension, as the full project is intended to push the bounds of whatever these AI are. Whether that be parrots or... something more. I'm not promising to give out the architecture, but giving out the means to AGI would be the nice thing to do, I suppose. Still doesn't mean I will, I don't want that misread. So if that's an issue, it's ok not to donate.

What are this project's goals? How will you achieve them?

Developing the new architecture, training/raising multiple models under means not considered in the industry. First virtual embodiment would be likely, then physical. How? By... doing these things lol?

How will this funding be used?

For testing, developing, etc., I'm not getting supercomputers but it's still some money for decent horsepower. Robotics are another large part of the budget. Plus things I can't do myself, like properly designed and rigged 3d models, which will be unique to each... I am making anywhere between 2 to 5 AI of my architecture, and not just because I'd like to, but because it may help in the learning process. the top end of the normal budget is a little above the minimum set amount since I don't have to get perfect hardware for compute (fuck matrix multiplication lololol) but the overall funding goal is to research the extent of this architecture and for more advanced robotics. gh200 systems, finding someone like BD for custom robotics that didn't come out of someone's hobby shop. that kinda vibe. I'll go as far as I'm allowed.

What are the most likely causes and outcomes if this project fails?

The outcome will be that nothing happens. I will not be able to do this without external funding.

What other funding are you or your project getting?

absolutely jack.