AI chat bots’ lack of conviction

quote:

---

Are they just telling us what we want to hear?

---

That is exactly what they are trained to do unless you run into a topic that where it is specifically instructed not to. If you can’t help injecting your own bias into a chat bot, they won’t be very useful.

There's a setting in Grok to change how it responds, the "inquisitive" setting I chose makes it more work than it's worth to get what should be a black and white answer.

The uncertainty may be a "defense" coded in, to try to combat the tards that just read whatever AI spits out and accept as gospel. Like the thread on the OT where someone asked ChatGPT what Frontier's check in policy was, instead of just going to Frontier.com

quote:

---

Messing around with Grok and ChatGPT I always notice how it gives an answer, then I ask it “but did you consider” and then it gives in way too easily. It seems even weak counter arguments will make it change its mind. Are they just telling us what we want to hear?

---

When you realize that they don't have a "mind" to change (and no real conviction either) they become a lot more useful. If you are arguing with it, it seems like you are asking for an opinion, in which case of course you can make it lean one way or the other. Yes, it is giving you the answer you want to hear.

These are tools. They are not "smart". They are really good at arranging words and creating images.

Give them better prompts. If you ask a simple question in a simple way, you will get a simple answer. If you ask for detail, you will get detail. If you have an opinion that you want reinforced, you will get reinforcement. If you have an opinion that you want it to argue against, you will get an argument against it.

If you think you have a good idea, don't ask if it's a good idea. Tell it to play devil's advocate from the start and tear your idea down. Use it as a filter to avoid making obvious mistakes or wasting time.

AI is a tool, the mind is yours to change or not.