ChatGPT is almost fixed now.

At first ChatGPT lied and said Dr. Poi Sonai wasn't real but after pressing it further, it told the truth.

I hope OpenAI keeps up the hard work.

31 points · 7 comments · view on lemmy.world

7 Comments

Rentlar@lemmy.ca · 15 pts · 4d (1 reply)

[Yes or No Question]?

[No/Yes].

That's not right.

My bad. [Yes/No].

My AI has a 100% success rate for Yes or No questions within 2 tries, using the special prompting method I devised!

Shows you how useless AI is.

Z745812939054@lemmy.zip · 15 pts · 4d (3 replies)

for almost any given output, you can reply "that's not correct," or something similar, and AI will agree with you

You can also tell it to search for the answer and provide links to the primary sources.

Naich@piefed.world · 2 pts · 3d (1 reply)

That's not actually true.

You are correct, good catch! It is not actually true that you can reply “that’s not correct,” or something similar, and AI will agree with you.

Would you like for me to explain how AI works, and how I generate my responses?

OpenRacism