One of the primary issues we (humanity) have been persistently grappling with is the need to have access to resources. “Where are the resources?”, we say, and procure vast resource repositories for a specific audience. Some within the niche know that the resource exists, others do not. Provided the resource is accessible, introducing a Large Language Model that works within a specific niche may truly increase access to resources in a broadly impactful way. Then we’re stuck with classic issues of access. What does this actually look like though?
- Pair the abilities of LLMs with niche resource repositories.
- Pair the LLM with an interface to refer to one or many repositories.
- (Optional and nefarious) Charge the user to access each repository.
This is surely the next step as we hear about the privatization of data being used throughout different projects (Bloomberg, LexisNexis, etc), along with a newfound desire to keep information private going forward (podcast link when I find it).
Beyond the clear commercial desire, the ability to introduce an LLM to navigate resource repositories serves as a clear opportunity to enable the general public to access information that is specific to their niche. By that I mean the public has the potential for increased agency as it relates to information access. In the past few years, many of us have become familiar with the concept of an everything app or a “super app” because the worlds richest are grabbing for chips and building what they can (it feels like Forbes is minting more million and billionaires), or because of their proliferation everywhere but the US.
What has prevented them from making headway though? I have my own theories, but I believe that they will start to look more attractive as their usage makes more sense. Those who have the opportunity and ability to experiment with ChatGPT now will forge the path for those to follow. As people and businesses begin to understand how they can implement their own LLMs, microcosms of niche repositories will be created and enabled by AI tools like ChatGPT.
Again though, how do we get there? Text-based input is of course the norm for processing information, but I was listening a podcast host who equated natural language input to being able to cast a spell (I will cite as soon as I remember). In a simplified way: natural language input followed by complex output. The input could be language, text, images, scents… textures? It all comes down to interpretation of the data.
The ability to work with AI tools will continue to trend towards easier methods of “spell casting” on whatever it is you’re working with. You’ll be able to parse content you’re engaging with through the AI, which will in turn produce the “spell”.
[ input -> output ]
As “spell casting”/natural language input/prompting becomes easier and more nuanced, more people will have access to casting spells. There will be different spell manufacturers, makes, and models, but the general trend of input -> output will continue.
As I’ve previously discussed, I believe small businesses are the primary benefactors in this scenario. In terms of what was possible, and what is now possible, a large or medium-sized business is not making gains as substantial as a small business would. Most larger businesses would focus on implementing cost-cutting measures, and quite possibly removing employees or contractors from the situation. A truly small business would likely be additive as they are implementing processes into their workflow that may not have been possible before. Those larger businesses that are able to implement additive strategies without detracting from their bread and butter would be those to keep an eye on.
When I first started talking about super apps five years ago, I felt mad. An all-in-one app that connects all of my services seamlessly? Typing this out it still feels like a pipedream, at least in the US, but the reality is closer than we think. We’ve seen clear example overseas in apps like WeChat or Grab, but imagining a version at home is more of an exercise. Perhaps this is why he-who-must-not-be-named is pushing the everything app so fervently. The need to have the concept demonstrated will give way to others believing that they too can push their own version.
Where are the resources? Right under our noses.