Yes, if you’re a programmer. A lot of real-world business scenarios are basically a non-technical person using English to give a (flawed, incomplete) specification, and a programmer using their expertise not only to implement but also to fill in the incomplete parts of the specification and also to determine what missing parts of the specification can’t be interpolated and have to be asked about. That sort of filling in missing parts based on precedent is the exact sort of thing these AI are good at.

Jeff

On Mar 31, 2023, at 4:27 PM, Mig Mit <migmit@gmail.com> wrote:

How is it going to benefit anyone though? English, like all natural languages, is terrible for describing algorithms. It is way easier to just code the damn thing.

Sent from my iPad

On 2023. Apr 1., at 1:24, Yasuaki Kudo <yasu@yasuaki.com> wrote:


There is TTS (text to speech) and STT (speech to text)  Former is easier, latter is harder.

Similarly, is it possible that we generate a huge variations of English explanation of correct compiled programs (like TTS - easy-ish, because we can program English explanations algorithmically ) and the get GPT-like-things to learn from that huge sample so it can convert human instructions to accurate programs? (like STT - hard),


On Apr 1, 2023, at 08:08, Branimir Maksimovic <branimir.maksimovic@gmail.com> wrote:

Of course it’ useless. GPT should be tested with something that can’t be found
on internet :P

I tried with problem euler 512, it can’t find optimal solution, but can perform
microoptimisations of existing solution, which is still impressive :P

Greets, Branimir.
 

On 1. 4. 2023., at 01:01, Alexey Vagarenko <vagarenko@gmail.com> wrote:

Isn’t binary search useless for lists? It performs worse than full list traversal.

Сб, 1 апр. 2023 г. в 03:43, Branimir Maksimovic <branimir.maksimovic@gmail.com>:
I was thoughtt from my mathematic logic proffesor in 1987 that algorithm for making algorithms
doesn’t exist. He proved that there is no algorihm for provin hieor order logic formulas…

Greets, Branmir.


On 31. 3. 2023., at 23:37, Yasuaki Kudo <yasu@yasuaki.com> wrote:

I am really curious - is it possible to create a programming language designed from the ground up to be used by GPT and similar tools?

- Easy to specify the desired outcome
- Easy to verify results

I tried GPT4 to produce COQ proofs of Yoneda Lemma but it could not even complete the interfaces (without implementation) - they failed to compile, by being tripped up by implicit parameters etc.

But GPT4 gets much closer to completion compared to GPT3.5

I wonder what "GPT-native" programming language might look like😄

-Yasu



On Apr 1, 2023, at 06:24, Levent Erkok <erkokl@gmail.com> wrote:

This code is buggy when lo + hi overflows your integer type. Classic (extremely) corner  case bug: https://ai.googleblog.com/2006/06/extra-extra-read-all-about-it-nearly.html

On Mar 31, 2023, at 1:02 PM, Gregory Guthrie <guthrie@miu.edu> wrote:

From a colleague:
  GPT Pretty smart(?!).
But this is of course simple, and searchable.
 
Still…
--------------------------------------------------------------------------------------------------

GPT-4 when prompted:

Write binary search in Haskel. Print results using putStrLn. Ensure bug-free code by specifying preconditions, postconditions, invariants, variants, internal states.
 

gives the following response:

-- Binary search in Haskell
-- Precondition: the list is sorted in ascending order
-- Postcondition: the function returns the index of the target element if found, or Nothing otherwise
-- Invariant: the search space is always within the bounds of the list
-- Variant: the search space decreases by half at each recursive call
-- Internal state: the lower and upper bounds of the search space

binarySearch :: Ord a => [a] -> a -> Maybe Int
binarySearch xs x = go xs x 0 (length xs - 1)
  where
    go xs x lo hi
      | lo > hi = Nothing -- base case: search space is empty
      | x == midElem = Just mid -- base case: target element is found
      | x < midElem = go xs x lo (mid - 1) -- recursive case: search in the left half
      | otherwise = go xs x (mid + 1) hi -- recursive case: search in the right half
      where
        mid = (lo + hi) `div` 2 -- calculate the middle index
        midElem = xs !! mid -- get the middle element

-- Print results using putStrLn
main :: IO ()
main = do
  let xs = [1, 3, 5, 7, 9] -- a sorted list of integers
  let x = 5 -- a target element to search for
  let result = binarySearch xs x -- call the binary search function
  case result of -- pattern match on the result
    Nothing -> putStrLn "Element not found" -- print a message if Nothing is returned
    Just i -> putStrLn $ "Element found at index " ++ show i -- print a message with the index if Just i is returned
 

The above code runs:

 
 
<image001.png>
 
_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.

_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.
_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.

_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.
_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.

_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.
_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.
_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.