Does generative AI facilitate investor trading? Evidence from ChatGPT outages

Post by 01HNNWZ0MV43FF » Wed, Jul 17, 2024, 9:09 PM UTC

dorkwood wrote: post #38

ProllyInfamous wrote: post #26

Background: I am personally a long-term value investor ( a la Sir Templeton or Buffet †), but annually breakfast with a retired "numbers guy" (who's currently building his short-term trading strategy). For decades he worked finance for among the largest accounting firms in the world... Since 2016, our mutual friend and I have been adamently suggesting he look into "this bitcoin thing" — his response was typically [di…

> His response was simply "I never got embarrassed asking the computer any finance questions, even stupid/simple ones; I've always been the finance guy with answers so I forgot how to be humble in unexplored topics." I think I'm the opposite. My feeling is that anything I type into ChatGPT gets stored in a database, ready to be leaked at some future date, or read by prying eyes without my knowledge. I'm more careful…

Post by Arn_Thor » Wed, Jul 17, 2024, 9:11 PM UTC

joering2 wrote: post #8

I think lately results of chat went downhill. I got used to asking "are you sure about that" after each chat answer to my question, and to most of the time the answer is "you are right, here is how I was wrong".

Post by adw » Wed, Jul 17, 2024, 9:16 PM UTC

Workaccount2 wrote: post #31

ec109685 wrote: post #24

Earlier quoted context omitted.

Ask a question like “Why do EVs use less electricity per mile when it is 100 than 70 degrees outside” and then reverse the question. You’ll get two answers that totally contradict each other. “Electric vehicles (EVs) use less electricity per mile in warmer temperatures (around 100 degrees) compared to cooler temperatures (around 70 degrees) primarily due to the following reasons: 1. *Battery Efficiency*: EV batteries…

This type of behavior seems like it should be something that should be (relatively) easy to fix with some kind of external intervention layer. Google has been able to answer queries without using LLMs for some time now, so maybe some kind of pre-check followed by a prompt injection to stiff-arm the LLM in the right direction.

Post by wongarsu » Wed, Jul 17, 2024, 9:27 PM UTC

01HNNWZ0MV43FF wrote: post #64

kjkjadksj wrote: post #48

Earlier quoted context omitted.

Time in the markets beats timing the market if you actually have capital to throw around to gleam a decent 5% yearly return. If you don't have much capital, you have to time the market like hell to build up your capital to a level where 5% yoy actually is a decent amount of money.

If timing the market works at a small scale, why stop? If it worked, wouldn't everyone do it?

Post by arrowsmith » Wed, Jul 17, 2024, 9:45 PM UTC

adw wrote: post #65

Workaccount2 wrote: post #31

Earlier quoted context omitted.

This type of behavior seems like it should be something that should be (relatively) easy to fix with some kind of external intervention layer. Google has been able to answer queries without using LLMs for some time now, so maybe some kind of pre-check followed by a prompt injection to stiff-arm the LLM in the right direction.

Whenever you're building some ad-hoc non-differentiable circumstantial hacks around your learned model (regardless of whether it's an LLM or not), two alarms should go off in your head; a) Could this be made differentiable? End-to-end systems are nearly always stronger than systems made up of multiple models or models-plus-heuristics; b) is it appropriate to be using a statistical model here at all? External interven…