KuriousCat

I an any business deal or negotiation, both parties negotiate and come to mutually agreeable terms. In case of software or digital purchases, it seems the providers can unilaterally impose terms, change them as they deem fit even when the user does not want to be bothered by the updated terms. It is even more pressing when it comes to AI models, the companies can scrape and use all the content available for them but explicitly write in their terms of agreements users can’t use the output train their systems. How did we come to this?

There is a strong tendency to integrate LLMs with existing apps. Barring few errors that could eventually get corrected, LLMs are fundamentally programmable computational environments. With that notion, every app could be an environment that is dynamically instantiated. Do we need to fundamentally change software delivery? Treat LLMs as dynamic computational environments, generate purpose-specific interfaces on-demand, create adaptive, context-aware interaction models. Eliminate fixed application paradigms?

AI companies are currently training their models using all the data that they could find, much is being debated whether data uploaded into the private repositories remains untouched. For instance, are git private repositories currently being used? Or your data on cloud storage drives is being used? Or device users are assumed to opt-in and data in edge devices is pulled in. Keeping this trend in mind are cloud free companies that never trusted the cloud services going to be most valuable in a couple of years?

How do LLM servers handle contextual data? Is the context passed on a prefix to a stateless machine? (Would mean a lot of tokens have to reprocessed during a session) Or a separate LLM instance is created and maintained for an active session? (Expensive and inefficient)