The following's a summary of my article I've Invented a New AI Algorithm. Given it turned out to be over 40 pages long, I think a summary might help people who aren't interested enough to read that much. I do want to say - by reading this, rather than the full piece, you'll lose a lot of details, several important points, not see a few demos and lose a lot of the nuance. But I get it - your time is extremely valuable, and I don't want to take it up with something that might not appeal to you.

Introduction + "Myself & SPC"

Hi, my name's Joseph. I've invented a brand new AI algorithm from first-principles, completely on my own, that I call "SPC". It's the first new AI algorithm since 1957, and it's a potential competitor against artificial neural networks (ANNs). It's the first of several new AI algorithms I possess, based on my private, self-funded, independent research over the last 13 years.

SPC

SPC was invented originally in December 2018, reduced to a working system between February and March 2024. I'm currently filing patents, set to be done in a few weeks, and so I can't freely talk about technical specifics until then. However, I can walk through key ways SPCs are different from ANNs. And when I do release technical specifics in mid-August, they'll be accompanied by a large amount of educational materials and a few software implementation libraries.

• They can learn with very small amounts of data.
• They are performant on local hardware. My development was done on an 2012 computer running Windows 7.
• They are more flexible than ANNs, because you can separate and combine them on the fly and train each layer separately.
• SPCs can provide full rights attribution to training data, while ANNs can't.

SPC Technologies

In this section in the main article I also discussed several key motivators for myself, some goals I have and tooling I'm building which may be of some interest to various readers.

There's a lot of potential things you can build with SPC. I've been working on a few, and know there's plenty more. Going through the technology I've already prototyped:

• I have an SPC driven tool for up-scaling, de-noising and recoloring images.
• There's a tool for converting multiple images into 3D point-cloud / voxel spaces. There's also SPCs for working directly on 3D voxel spaces.
• You can use one of the image or voxel SPCs to program simulations from images.
• There's the potential for SPCs that can generate 3D rigs and meshes.
• You can use SPC inside of shaders. This should let us make AI driven shaders for movies or video games that can be trained on very little data.
• Last but not least, there's a way to use SPC to train bots to echo your behaviors in a generalized way with as little as one training sample. This is a paradigm shift for agentic AI that outperforms ANNs, and is fully deterministic (reliable). I talked a lot about this in the main article.

Please Help

I'm in need of help. I've been self-funded for 19 months, but around 13 months ago now I sustained a brain injury that vastly complicated things. I've been unable to secure investors, or find employment (120+ apps, 1 interview) in the interim. I've got around 5 weeks of funding left, before I'm screwed. If that happens, to protect my research, I'll be burning all 13 years of it - which will set back AI technology 13 years.

I'm looking for any opportunities. I can license SPC, or my software tools. I'd love to work with anyone in a collaborative manner to improve their products or tools with mine as well, or provide consultation on SPC or AI technology. I'm also looking for potential jobs, though I'd like if they were focused on AI, SPC, applying my technologies or tooling. If any investors are interested in funding, I need it on an extremely tight turn-around, but you would have an opportunity to get in on owning a core AI technology that may displace ANNs in several key applications, including agentic AI.

If you'd like to reach me for any reason, you can email me at joseph@ikisoftware.com, find me on Twitter at @ikishade, or BlueSky at ikishade.blusky.social. I've also setup a Patreon, if anyone would like to fund my research or help me out.

The Future of AI

In this section in the main article I discussed quite a lot in detail. It's worth checking out if this summary interests you at all.

My goals are lofty, and I know that. But I'm working towards something I call "Industrial Science", which I think can be accomplished in the next 20 years. This will have the potential to revolutionize human kind. To get there I have a 5-phase plan. Phase 1 is a focus on creating AI-driven augmentation technology. Phase 2 is the construction of substantial simulation systems in the form of an MMORPG to raise Strong AI candidate systems in. The Strong AI use several AI algorithms that are not publicly known, and do not use ANNs in any way. Phase 3 is extracting the Strong AI ("Halykons") from the MMO-system, and using one of the other AI algorithms I have to create a group that have strong symbiotic relations with humans that can run on completely local systems ("Galateans").

For various reasons, I've decided not to discuss phase 4 and phase 5 in any detail right now, but once they've been completed, we will have a system that can industrialize the scientific process and result in accelerating the rate of human technological progress from somewhere between 5x to 20x. I want to aim this system at solving the scientific development of consciousness uploading, which will mean that the majority of those below the age of 60 at present may have a strong chance at post-corporeal existence. I also believe once we complete that, I'm done - but others could revolutionize how we explore space by swapping from an organic-astronaut centered mindset to a robotic-astronaut centered mindset.

The Field of AI - Part 1

But the field of AI is riffe with problems right now. There is the issue with Stable-Diffusion (SD) based AI art, and how it's created a highly unethical wealth transfer. There's the issue of people using AI driven unemployment to raise their stock-prices when it seems more likely that political factors (section-174) are more likely to blame for the 2023 to present issues. There's people cheering about unemployment, despite known correlations to suicide - which is fucking shitty and pisses me off.

I have issues with how much pseudoscience is in this field, to the point you could read 300 papers and find 1 saying anything real. AGI is laughably misunderstood, because it's categorically incapable of consciousness. We've struggled to even define intelligence, but people want to say we've almost reached super-intelligence. This is all a joke.

A joke that's motivated by a gluttonous appetite for money. I don't have issue with wanting money, but right now antisocial gluttony is causing lots of social problems. I want to work with the field of AI to ensure we are responsible, as I bring us into a new technological age with my research. At the same time - if people won't behave, I'll gladly use my AI in an adversarial manner to fight other AI technologies. SPC could just as easily be used to poison every training data set on the planet, including textual, code, music and image. And that's one of many techniques I have, and nobody else knows anything about the others. But I prefer collaboration over conflict.

The Field of AI - Part 2

I didn't want to end on that note however. There are several other cool things I could leave off on instead. So let me list them out quickly.

• All AI up until now has been relegated to 3 "AI domains". SPC is in a 4th. I'm aware of more than 12+ at present. By my estimates, all AI research done by humankind until now covers less than 10% of the field's breadth, and "Strong AI" is no where near the end of the field. It's more like going from a horse to a Model-T, and there's stuff in the distance on the level of space stations and ultrasonic jets. So Strong AI isn't particularly significant in the grand scheme of things, outside of how significant it is to humans because of how it marks the end of the anthrodominic age.
• There's a difference between AI techniques that can become sentient ("Ghost AI") and those which can't ("Doll AI"). Most AI researchers today can't tell the difference, which speaks to systemically insufficient education and debases the value of a doctorate in this field. None of the current AI techniques known to the public could be classified as "Ghost AI" without extreme modifications. So most people have gotten spooked by dolls.
• When we get more AI primitives, there's going to need to be a new programming language that has these primitives in it. We don't need this yet, but in a few years when me or other's release more algorithms, it will become necessary. For good reasons, this lang should be based on C++.
• There's an unexplored non-AI programming paradigm: ecological code. We could start building this today. If you want to learn more about this, I wrote some more in the main article. But it's effectively programming inspired by ecological systems in nature to create emergent, self-maintining systems.

Conclusion

And with that, you've got the summary of the article. Thank you for your time and consideration. I hope you have a great day! Cya!