Hacker News Reader

Comment by micheljansen | Hacker News Reader

This article reminds me a bit about a core theme in Isaac Asimov's Foundation series, which I'm coincidentally reading (mild spoiler warning). The science of psychohistory that is used to predict the future and create the Seldon Plan, only works if the future is allowed to unfold uninformed by psychohistory itself. That's why the Foundation was founded without any psychologists or knowledge of the Seldon plan. Seldon's predictions are based on the behaviour of all human of the galaxy, but knowledge of those predictions changes the outcomes, so they are no longer reliable.

As soon as AI starts participating in our world, the data that it was trained on from before AI was participating is no longer representative of the status quo. The very presence of AI changes the world it operates in.

View on HN

That's a great way of framing it. It's my expectation that we will ruin the internet as a useful training corpus by flooding it with generated articles, and we will end up with a pre-AI date we use to filter incoming data in order to avoid them.

I wouldn't be surprised if filtering regular, pre LLM bot spam was already a massive hurdle when collating data for ChatGPT.

Reminds me a bit of how mathematics gets into all sorts of trouble once self-reference / recursion is introduced.

Recursion is great and mathematics is doing just fine, thank you.

I think that's because trouble is the fun part

That's similar to what some faithful say when you criticize stock market technical analysis. That it only predicts when participants haven't seen the chart.

That is a really good insight.

George Hotz was making a good similar point of modeling self contained systems vs. open systems - a very different thing.

> knowledge of those predictions changes the outcomes, so they are no longer reliable

does the series explore whether the outcomes change for the better, for the worse, or simply for the different? curious to read them, I'm thinking about writing something similar, but with positive consequences for future knowledge

The whole idea of the series is to change future history for the better. Sheldon promised a thousand years dark age instead of a thirty thousands. SPOLIERS.

in the end it turned out that mind control was far more reliable and the foundation turned to more to mind control instead of subtly influencing history.

I’ve done this though experiment myself and is actually one of the issues I have with the concept of a rapidly self reclusive improving system. It would probably not be easy to stay ahead of yourself if the word your operating in changes from hour to hour ?

The funny thing is that probably the old wall street saying of “past performance is not applicable to future results” holds true even in this domain.

Reply on news.ycombinator.com