Hi, I'm Joseph Juma, the founder and currently sole member of Iki Software ("e-key"). I've invented a brand new AI algorithm, which I refer to as "SPC". This algorithm serves as a competitor to the two currently existing methods in the field of AI: artificial neural networks (ANN) and genetic algorithms (GA). If you're unfamiliar, almost all of the technology referred to as "AI" invented in the last ten to fifteen yearsDeep learning, machine-learning algorithms, transformers, encoders, decoders, autoencoders, stable diffusion (commonly, "AI image generators"), audio synthesis AI, large language models (LLMs, such as ChatGPT, Grok and Gemini), and so on are all based on ANNs. are driven by ANNs, while GAs are often neglected. The last time that a new AI algorithm with practical applications was invented was genetic algorithms 68 years agoYou could debate the last one was Douglas Lenat's work on symbolic AI in 1976, but that is categorically different from ANNs, GAs and SPCs. Even then, that would still only bring 68 down to 49., in 1957, and 13 years prior, neural networks were first implemented in 1944Technically the 1944 method is based on a work from 1943, and the first implemented ANN wasn't until the Perceptron in 1957..

Because of patent paperwork, I can't freely discuss SPC today. However, once I can (in mid-August) I'll release a follow up including software libraries and educational documents. So what are some features of SPC?

• It can be trained on very few samples, and is still effectve.
• It's able to be parallelized, so it works better on GPUs.
• It can have full rights attribution, and training data can be added and removed without issue.
• It's far more flexible than ANNs, and able to be dynamically reconfigured.
• There are 3 components you can tune to alter its behavior, similar to Cost and Activiation functions in ANNs.

You can see below an image of a tool I've built for up-scaling heuristic images. This runs on 1 training sample.

An example of an SPC being used to upscale a heuristic image of an island from 16x16 to 512x512 using one image worth of training data.

I've employed it already in a number of applications, including video game development, developer automation and data analysis. All together there are over 8 different technologies I've been focusing on employing it, but there are easily plenty more to come. If you are interested in learning more about SPC, want to gain early access, wish to license it, or discuss business partnerships regarding it, please feel free to email me at joseph@ikisoftware.com. Thank you and have a great day.

- Joseph