[Haskell-cafe] Bing/Copilot Answer: "->", has highest precedence for type construction but in general has nuanced precedence order

[jo at durchholz.org]

On 16.03.24 23:57, John Creighton wrote:
> In the summary part of bing's answer, bing actually said 
> that "->" has the highest order of precedence but if you read the whole 
> answer you see it is only for type construction where Bing 
> specifically made claims of the order of precedence and in the other 
> case (parsing & fixity), bing said that the fixity of "->" is not 
> defined in the Haskell 98 report, and nuanced it by saying that the 
> parsing rules depend on both fixity and associativity.

The thing with GPT AIs like what's currently hyped is that it's 
essentially just piecing together sentences on the basis of "in all the 
training data, what was the most likely follow-up word". IN particular, 
they do not follow any reasoning or do any semantic analysis whatsoever.

This is great for writing speeches.
It's also great for boilerplate code.

It's bad for accuracy. It essentially depends on the quality of the 
training data, and since these GPT bots need humongous amounts of 
training, that's typically "just the Internet". (The companies behind 
these bots try to filter out nonsense, with a focus on avoiding public 
outcry, which tends to create even more humorous results.)

Regards,
Jo