Comment by srcreigh
8 hours ago
LLMs receive new data via input context, not just training data.
Thought experiment: How effective will 2026 LLMs be for humans in 2526?
It's not game over just because 500 years are missing from the training data. The important question is how well can 2526 humans make culture and knowledge navigable to LLMs via tool calls.
Today's LLMs might need for example sub agents to translate to 2526 English, sub agents to read 2526 docs.
It's _really not clear_ whether 2026 LLMs will be useless. To believe that reflects an enormous misunderstanding.
> LLMs receive new data via input context, not just training data.
Be more specific about the "new data". If everyone is using LLMs for work (generating code), especially the juniors who won't get the chance to learn from first principles, LLMs will be training on the data they generated. How will new code enter the system at large enough quantity that it can be used for training?
> It's _really not clear_ whether 2026 LLMs will be useless. To believe that reflects an enormous misunderstanding.
They won't be useless, they will just be frozen knowing only whats in their training data. No new programming languages will emerge, in 2526 they'll still be using Rust and javascript, same exact code from 2022 which dominates the training data.
The "new data" is: person A prompts an LLM to create or modify a tool, person A distributes code person B, person B's LLM uses the tool via docs/help/error. That is a direct path for an LLM to "know more" from a human than what's in its training data.
If we get a new programming language not in the training dataset, we could give an LLM a decent compiler with compile errors, and some sample code and it would be able to write code in the new language without training.