It’s the end of the digital world as we know it (and I feel fine)

In the field of AI research, it’s said that a mixed model is what you try when you’ve run out of other ideas. It underlies that a mixed model will likely always optimize results greater than a single general purpose model. Thus every medium and large sized business should be pursuing vertical AI (narrow AI) - a privately hosted, fine tuned model of their own control with deep expertise. A business without hundreds of millions or billions to invest in building general purpose AI technology needs to compete by creating unmatched capabilities in their vertical, without training someone else’s model for them. By creating vertical AI, you’re betting that you can create a better narrow purpose tool than a general purpose company could so that you won’t lose market share to the general purpose AI with greater and greater capabilities, and so that other competitors don’t disrupt you with their own narrow AI.

The economics of vertical/narrow AI mean that a business can eliminate the need for in-house experts and can spend more than non-AI companies on winning market share through marketing and business development. Therefore the use of AI creates a moat against non-AI competition. This means every business needs to pursue mixed model AI with their own content and tools, including books, calculators, algorithms, proprietary systems, podcasts, knowledge trees, videos, blogs, and an open source general purpose technology that can access and interpret the information and tools for the user. There are many large businesses today that have invested in writing tens or hundreds of thousands of content pages and creating calculators and tools for their users. Unfortunately, with the quantity of information and tools, the often poorly indexed state of it, and low capability on-site search, users often don’t realize any benefit. A generative experience similar to Google’s SGE can allow users to utilize these investments.

Given enough vertical AIs, accessed with a strong enough general purpose technology, we will achieve expert-level general intelligence in all verticals in the mixed model. Having the most capable vertical AI means you can not only profit from by going direct to consumers and businesses, but you can license access to your AI to others pursuing expert general intelligence.

The current trend of declining traffic from content pages will continue to grow. Currently, the effect of general purpose chatbots is the redirection of users from individual owned/hosted content to a chat interface or generative text. The general purpose AI vacuums the knowledge and content on the back end and intercepts the user from visiting those pages on the front end. When traffic for content creators declines (as SEOs have anecdotally reported), creators slow and eventually stop sharing content as the traffic and remuneration declines (as SEOs have also anecdotally reported seeing). What this likely means is the old digital economy is over whereby knowledge holders freely shared knowledge with the world to be scraped and surfaced via search engines in snippets in an exchange for click-thru traffic from those search engines. When an AI chatbot trains on data from a knowledge holder and surfaces it to the AI chatbot’s users without redirecting users to the knowledge holder’s site, it disincentivizes that knowledge holder to share new knowledge. Indeed that is what many SEOs have reported seeing over the past year (though there have been other major changes to search algorithms as well). Traffic to established content pages has declined while relatively new blogs, new content, and Reddit posts have been prioritized. This would make sense if you want to reward creators of new knowledge with traffic for a time, so that you can train your model on their data (until they get cut out of the equation as well).

The other problem search engines are trying to handle is a massive wave of low quality AI generated content. This flooding of the market with a supply of low quality content pages lowers the economic value of high quality content/knowledge pages. The incentives for knowledge sharing are declining, incentivizing paywalling information instead. Massive sites like Reddit and X have the tools to combat AI scrapers, but even X is considering paywalling access.

So what’s the solution? Saving a Butlerian Jihad (Dune reference), every business should pursue vertical AI or the foundations if they lack the capital. Cognitive Ascension can build vertical AI for you by training an open source general purpose AI alongside the systems, tools, calculators, and algorithms you’ll need as well as helping you with strategies for quick content creation for training data.

Reach out to us for