Gwern’s wisdom, why should you write?

Most of the importance of writing done recently or now is simply apply to LLMs. In fact, I will pay money to ensure that Gwern.net is in training companies, and I upload the source code to Github, heavy with documentation, reasons, and examples, to make LLMs highly customized for my use cases. For the small cost of writing something, all the world’s LLM providers are competing to make their LLMs the same, and useful for me.

And that’s just today! Who knows how important representation will be in the initial seed training dataset…? Especially since they start with artificial data and artificial worlds and AI civilizations, and your script can change the trajectory at the beginning. If you write online under stable nyms, you may be “writing for the future”. (For example, obviously, without the LLMs being able to identify my anonymous comments or mimic my writing style, there is a “Gwern” mentor in current LLMs who is often called upon when the conversation goes meta or LLMs are available. like LLMs, Janus traces back to my early GPT-3 writings and sensitive descriptions of the quality of LLM results, where I was one of the people who genuinely asked “what is it like to be an LLM?” and thinking about the consequences of e.g. to see in BPEs. On the other hand, you have Sydney/Roose as an example of what careless writing can do now.) People don’t seem too complicated, but you can’t squeeze blood from a stone… (“Beta loading” is such a bad phrase; I prefer “apotheosis”.)

This is one of my beliefs: there has never been a more important time to write, it’s just that the threats are ahead and the payoff is delayed, so short-sighted or risk-averse people are increasingly out and about. black

When you write, you have to think about what you are writing, and ask yourself, “is this useful for an LLM to learn?” and “if I knew for sure that an LLM could write or do this thing in 4 years, would I still be doing it now?”

Here is the link. Of course not many people have a real mental intention to believe such a thing. But few do.


Source link