[raspyberr]

Here is an example of ChatGPT being confidently wrong in something domain specific that I know about. Here's what I asked:

"I'm playing Old School Runescape. I want to efficiently level up my Mining. I'm currently level 90 and I want to do a tick-manipulating method. Something like 3t4g. I don't understand when to click and what I need. Can you explain the method?"

Here's the method it replied minus kruft at the start and end: " 1. Equip your pickaxe and head to the location where you want to mine.

2. Find a 3x3 square of rocks that you want to mine. The square should be in a straight line in front of you, with the rock you want to mine in the middle.

3. Stand in the square so that your character is in the middle of the square, facing the rock you want to mine.

4. Start mining the rock as normal.

5. As soon as you see the animation of your pickaxe hitting the rock, immediately right-click on the rock and select "Prospect". This will cancel the animation of your pickaxe hitting the rock and allow you to perform another action.

6. As soon as you see the prospect animation start, immediately right-click on the rock and select "Mine". This will cancel the prospect animation and allow you to perform another action.

7. Repeat steps 5 and 6 as quickly as possible.

8. Every fourth action, switch to a different rock in the 3x3 square. "

This is total nonsense. Not only is this not 3t4g but the actions it says don't work like that in the game. Obviously it's not understanding anything and it's forming sentences that sound natural in the context of the question. OSRS has one of the best most complete wikis of any game. To me this is an example of being very cautious of trusting what it says when you ask it about something you don't know about.

[B1FF_PSUVM]

ChatGPT is a bullshit generator.

Earlier this week I was discussing with family why "you should not freeze things twice". Someone made the mistake of asking ChatGPT, and it confidently pooped out that bacteria keep reproducing in frozen food!

(The actual reason is that freezing does not actually kill bacteria already present - cooling/freezing stops the growth, but it is cumulative. So if you have stuff out, then re-freeze, then out again, the second time it may reach toxic levels.)

P.S. https://www.theregister.com/2023/03/02/chatgpt_considered_ha... - fake obituary link and all ...

[Angostura]

If you want a simpler example - try asking it for a list of english words where the letter V is silent.

On the other hand, I do find it quite useful for suggesting improvements to things. You can't rely on the suggestions, but they can be thought provoking.

[joshschreuder]

One of the words it gave me was “Yclept (archaic form of "called" or "named")” which is both silent and invisible!

[fallingknife]

> Obviously it's not understanding anything and it's forming sentences that sound natural in the context of the question

> To me this is an example of being very cautious of trusting what it says when you ask it about something you don't know about.

More and more like humans all the time

[worrycue]

Humans who don’t understand what they are talking about we should not listen to nor trust. By the same logic it would be stupid to listen to / trust ChatGPT.