Can GPT be Reliable?

[JohnH]

i would prefer to be allowed to speak freely, not have a priesthood controlling what we may write post or say based on whether it agrees with “science” etc

[Bootstrap]

i would prefer to be allowed to speak freely, not have a priesthood controlling what we may write post or say based on whether it agrees with “science” etc

You absolutely have that right. There is no controlling priesthood here.

[JohnL]

i would prefer to be allowed to speak freely, not have a priesthood controlling what we may write post or say based on whether it agrees with “science” etc

I agree with you. This demanding of proof or whatever is condescending at best interpretation and fascism at its most pessimistic interpretation. The Holy Spirit inspires intuition and insight and these are key gifts to understanding humans and their behaviors. I believe God laughs at arrogant men who try to control inputs and other things that must seem petty in his eyes. God is in control. No one else.

[Bootstrap]

i would prefer to be allowed to speak freely, not have a priesthood controlling what we may write post or say based on whether it agrees with “science” etc

I agree with you. This demanding of proof or whatever is condescending at best interpretation and fascism at its most pessimistic interpretation. The Holy Spirit inspires intuition and insight and these are key gifts to understanding humans and their behaviors. I believe God laughs at arrogant men who try to control inputs and other things that must seem petty in his eyes. God is in control. No one else.

I agree with the last part - God is in control.

And there are no demands of proof. But if you bring up a complicated subject that is a matter of fact, and facts are available, there is also no controlling priesthood saying those facts must not be brought up. It cuts both ways.

I'm not going to stop explaining how GPT actually works just because it annoys you. The truth that people don't want to hear is often a truth that should be spoken. And facts can help sort out what is true. Feel free to provide your own facts any time. There is also fact-free arrogance. And tearing down people rather than discussing subjects.

[Neto]

I thought of this Brazilian joke (from back in the mid 80's), and decided to see how well ChatGPT would do with translating it, and then explaining the punch line. It did amazingly well.

I wrote:

Translate this joke from Portuguese to English, and explain the punch line: Porque o Presidente Sarney mandou VW fazer o fusca de novo? Porque ele não sabia diriger Brasilia.

ChatGPT's response:

The joke in Portuguese is:

"Porque o presidente Sarne mandou a VW fazer o Fusca de novo? Porque ele não sabia dirigir Brasília."

The translation to English would be:

"Why did President Sarney tell VW to make the Beetle again? Because he didn't know how to drive Brasília."

Explanation:
The punchline plays on the double meaning of the word "Brasília" and the idea of a car model. Here's how it works:

"Brasília" is the name of both a car model (the Volkswagen Brasília, produced in Brazil) and the capital city of Brazil, which was designed by architect Oscar Niemeyer.
The joke sets up the idea that President Sarney, instead of knowing how to drive the Brasília car, didn’t know how to drive the city of Brasília (in the sense of managing or leading the country, since Brasília is the capital and symbolically the center of power).
The humor arises from the mix-up of the two meanings of "Brasília" and the implied absurdity of a president not knowing how to manage the country but instead being concerned about car models.

It only messed up in the last paragraph/sentence. There is no intended relationship between the car and the (former) president at all. It was a criticism of the president's governmental abilities. The joke is possible because the Portuguese verb 'dirigir' is used both in the sense of driving a car, as well as leading an entity, in this case, the government of Brazil. (Brasilia is the federal capitol of the country.) And yes, he actually did order VW to restart manufacture of the VW 'bug', called the Fusca in Brazil. His reason was to restore the availability of the most economical vehicle then in use. The military had just relinquished their control over the government, after appointing presidents out of their own ranks for around 15 years. President Sarney was not elected - the President elect had surgery for appendicitis there in Brazil, and died from infection. He intention in not going out of the country for the surgery was to build confidence in medical care in the country. It apparently did not live up to his level of confidence. So his appointed vice president, Jose Sarney, took his place. His instituted price freezes on virtually every product. The years of his term and those that followed were a time of uncontrolled inflation, like around 1,000% some months. The government knocked 9 zeros off of the currency between 1985 and 1994 (three times, 3 zeros at a time), renaming the currency each time.

[Neto]

Here's something that popped up in the news today:

https://www.bbc.com/news/articles/c0m17d8827ko

AI chatbots unable to accurately summarise news, BBC finds

[Robert]

Here's something that popped up in the news today:

https://www.bbc.com/news/articles/c0m17d8827ko

AI chatbots unable to accurately summarise news, BBC finds

They did not test or include Grok.

[JohnL]

Here's something that popped up in the news today:

https://www.bbc.com/news/articles/c0m17d8827ko

AI chatbots unable to accurately summarise news, BBC finds

Neither can mainstream legacy media.

[Neto]

My wife & I were discussing this at breakfast this morning, and it occurred to me that perhaps the 'failure' of these AI tools may also reflect on the quality of writing that was fed into them. That is, it might be a worthwhile writing tool, just to check for clarity in writing. So if it makes unintended assumptions, then maybe it's the author's fault. The next step would then be to rewrite the piece so that it cannot be understood to convey these unwanted conclusions. (An interesting study would be to attempt to determine if the false assumptions the AI is making are ones that a human reader of the target audience might also easily make. The test readers could be given the article, then be asked some questions that reflect the AI's false assumptions.)

[Robert]

My wife & I were discussing this at breakfast this morning, and it occurred to me that perhaps the 'failure' of these AI tools may also reflect on the quality of writing that was fed into them. That is, it might be a worthwhile writing tool, just to check for clarity in writing. So if it makes unintended assumptions, then maybe it's the author's fault. The next step would then be to rewrite the piece so that it cannot be understood to convey these unwanted conclusions. (An interesting study would be to attempt to determine if the false assumptions the AI is making are ones that a human reader of the target audience might also easily make. The test readers could be given the article, then be asked some questions that reflect the AI's false assumptions.)

If they are being selective to left leaning sites like Google and a few others do, it could be causing some of the poor summarization. It could also be that it still needs to learn better ways of functioning. AI is still in its infancy. Has a lot to learn.

I am curious why they left Grok out of it.