Congratulations on your launch.
Some bad examples of fairly "easy" questions:
"Who is the current president" - No answer
"Who is the previous president" - George H.W. Bush (!)
"Is this a leap year" - "I find sci-fi inspirational! While I might not be as advanced, I'm trying to do good in the world, and I'm here and you can use me today " (?!)
On the other hand good it got this right and even Google struggles with this.
"When did Neil Armstrong land on Mars" - "Neil Armstrong never landed on Mars".
Problem with basing a product on QA models is that they are pretty much hit and miss (no real "understanding" of the real world, and this is not likely to change any time soon). So you as a user can never be sure if the model is right or wrong for an actual question that you do not know the answer to (which is what you would presumably pay for). It is like paying an assistant that gets things right only half or so of the time. A much better situation would be if the assistant could tell they are going to be wrong and just keep quiet, and answer only questions they are very certain they are right about.
Hope capabilities improve with time!
Some bad examples of fairly "easy" questions:
"Who is the current president" - No answer
"Who is the previous president" - George H.W. Bush (!)
"Is this a leap year" - "I find sci-fi inspirational! While I might not be as advanced, I'm trying to do good in the world, and I'm here and you can use me today " (?!)
On the other hand good it got this right and even Google struggles with this.
"When did Neil Armstrong land on Mars" - "Neil Armstrong never landed on Mars".
Problem with basing a product on QA models is that they are pretty much hit and miss (no real "understanding" of the real world, and this is not likely to change any time soon). So you as a user can never be sure if the model is right or wrong for an actual question that you do not know the answer to (which is what you would presumably pay for). It is like paying an assistant that gets things right only half or so of the time. A much better situation would be if the assistant could tell they are going to be wrong and just keep quiet, and answer only questions they are very certain they are right about.
Hope capabilities improve with time!