The Impossibility of LLM "Personalization" in the Two Dimensions of Text and Speech

Alleged "AI-first" User Interfaces are generally defined by no or buggy memory and session persistence. They are also avoiding the reality that "personalization" is impossible to manifest as valuable insight because there is no context in the current user interfaces that these, especially ChatGPT, use to interact with their users.

What does that mean?

ChatGPT announced "personalization" recently. But how do you achieve that? How do you personalize with only the vectors of speech and text, with no labelling of chat histories within chat sessions or ongoing sessions, without clearly outlining what category your interactions fall into?

What chats represent my base persona?

For example, how does ChatGPT know when your interactions are your base persona, the points of data and interactions that define you as you? As opposed to sessions that are playful Q&As about movies or topics that caught your interest on your X timeline or TV news. How does it differentiate your base persona from a chat session where you are working on the sci-fi novel that you are writing?

In other words, how does the lack of labelling and annotation allow the flat ChatGPT LLM to distinguish the base persona of "you" from these episodic and wide-ranging sessions and sessions within sessions on ChatGPT that relate to topics that are unrelated to your base persona but into which these data and "memories" stored by ChatGPT will bleed into and thus create a false or at the very least thoroughly inaccurate base persona of you.

It has to guess, and it is not good at guessing

It cannot. Therefore, it must either attempt to guess the base persona level interactions and, in doing so, make assumptions based on inferring rather than receiving clear, overt clarity and instruction from the user themselves.

Even if that interface were available to allow the user to guide ChatGPT on this path, who would spend all that time annotating and labeling every session, or chat history, or session within a session to achieve personalization, and for what return?

Who benefits from LLM personalization?

Well, by Web 2 standards, ChatGPT does so by selling that data to advertisers, market segmenters, and brokers. The migration of Web 2.0 data moats, unauthorized monetization, and unauthorized surveillance capitalism to narrow AI LLMs is simply the amplification of an already problematic dynamic in people's relationships with Big Tech.

It's time we changed that. That is one of the reasons why we built OS Mission at Open Code Mission. That is why the Verum Sphere exists within the Open Code Mission stack. That is why we deploy Strands, not applications from OS Mission, Strands that create Lumens, not data. Lumens because they are illuminated by explainability, interpretability, and ownership by the person, not the technology company.

Open Code Mission creates sovereign personalization

The iTRAXian multi-modal input-output layer that OS Mission uses to communicate with people, AI agents, and other sovereign AI ecosystems, built on the same principles, is how Open Code Mission delivers true sovereign personalization.

Open Code Mission delivers personalization, allowing the sovereign user to decide how to utilize it. Big Tech wants to misuse the personalization it delivers. Each member of the Open Code Mission ecosystem can choose to transact their self-sovereign data directly to external third parties in exchange for value.

They do not have to agree to the terms of service for the sole monopolistic and opaque benefits of Web 2 Big Tech behemoths' attempts to migrate their proven abuse of power and hegemony to AI, blockchain, and the new currencies that will run this new economy that is person-based, not anti-competitive monopolistic in nature and travels with flawed central bank fiscal policies.