[kwlug-disc] ChatGTP + Bing: we're entering a new paradigm and it's mind-blowing

[Doug Moen]

The Turing Test has been passed, and many people are disappointed by the shitty automaton that passes it.

The first time this happened was in the 1960's, with [the ELIZA program](https://en.wikipedia.org/wiki/ELIZA), which simulated a psychotherapist. Public reaction was weirdly similar.
> ELIZA's creator, Weizenbaum, regarded the program as a method to show the superficiality of communication between man and machine, but was surprised by the number of individuals who attributed human-like feelings to the computer program, including Weizenbaum's secretary. Many academics believed that the program would be able to positively influence the lives of many people, particularly those with psychological issues, and that it could aid doctors working on such patients' treatment.

ChatGPT is a bullshit generator, and it is *very* good at generating plausible and convincing bullshit. Sometimes it just spits out something it copied from the internet, just enough to engender trust, but in many other cases it "hallucinates" information that looks plausible and trustworthy, but is totally made up. What ChatGPT is optimized for is fooling humans.

I read a CACM article <https://cacm.acm.org/blogs/blog-cacm/268103-what-do-chatgpt-and-ai-based-automatic-program-generation-mean-for-the-future-of-software/fulltext> by Bertrand Meyer, a famous computer scientist, who was extremely impressed by ChatGPT's ability to write computer programs. The prompt was
> I have a function which gives 0 for 0, 1 for 1, 4 for 2, 9 for 3, 16 for 4, 25 for 5, and 35 for 6. Can you please give me the code in Eiffel?

The output was wrong, since the code used the formula `n * n`, which returns 36 for 6, not 35. Meyer explained why the program was incorrect, and asked for it to be fixed. The next attempt used a chain of if statements, and was correct.

Then Meyer asks
> Is there a more general way of phrasing it, rather than case-by-case

ChatGPT replies:
> Yes, there is a more general way of expressing the function that returns the desired output. Instead of using a series of 'elseif' clauses, you can use a formula to calculate the value for any input 'n'.
> 
> The function can be written as follows:
>    function foo(n: INTEGER): INTEGER
>    do
>       Result := (n * (n + 1) * (2 * n + 1)) div 6
>    end
> This function will return the same values as the previous version for the inputs 0, 1, 2, 3, 4, 5, and 6.

Meyers:
> The inferred function is rather impressive. What human would come up with that function in less time than it takes to say "Turing test"?
> We can only say "hats off". Should we also head for the employment office in search of mid-career retraining sessions for professions with a future?
> 

But the third program is bullshit. The outputs for [0,1,2,3,4,5,6] are [0,1,5,14,30,55,91]. Meyer didn't catch this and came off looking like a fool.

Because the answer to his question "What human would come up with that function..." is a scam artist who is trying to impress or manipulate you, who has no respect for the truth.

But we shouldn't anthromorphize ChatGPT. It isn't a human equivalent AI, it isn't a person, it's a clever chat bot built on top of a content associative database of shit people say on the internet.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://kwlug.org/pipermail/kwlug-disc_kwlug.org/attachments/20230215/6e77f6ff/attachment.htm>