Many major tech companies spend untold millions of dollars and run countless bots in an attempt to scrape every bit of data they can from the internet. All of that massive bulk of data then has to be processed and filtered, as companies attempt to reach a minimum threshold of quality.

Training a single LLM can take $10-150 million USD, but gathering and cleaning the data to feed that system is an additional expense.

In conventional LLMs, once a system has trained on data, that data is neither truly “remembered”, nor can it ever be fully removed. Consequently, the data also can’t be improved over time, as it was never really there to begin with, only a tangled mess of heuristics describing it is preserved.

AI Regulation is a hot topic on the table, that will remain on the table, imposing considerable long-term legal and financial risks to AI companies.

Writers, artists, and a growing number of other professions require viable ways to monetize their contributions within the rapidly evolving domain of Generative AI. To accomplish this in a fair and meaningful sense requires being able to separate data that was used at a given time from data that was not. While neural networks don’t offer this capacity, ICOM-based systems like Norn can.

With Norn’s capacities applied, credit may be accurately and proportionately assigned, while maintaining compliance with even strict privacy regulations, outperforming less ethical systems at a fraction of the cost, and using a tiny fraction of the data.

“Scraping” the bottom of the internet’s barrel is costly, hazardous, and completely avoidable.