Re:DeepMind’s AlphaCode Conquers Coding, Performing as Well ...
by nsxwolf on Thursday December 15, 2022 @01:58PM
(#34003987)
Tickets are rarely correct as written. Is the to AI going understand that and conduct meetings to hash out the missing details and corrections?
[ Parent
| on HN ]
Re:DeepMind’s AlphaCode Conquers Coding, Performing as Well ...
by ilaksh on Thursday December 15, 2022 @02:44PM
(#34004619)
It can send a chat message and hold a dialogue. For some small tasks things that is enough. Again main limitation is really vision and memory but we have to anticipate there will be very significant progress on those things in the next few years.
Re:DeepMind’s AlphaCode Conquers Coding, Performing as Well ...
by acuozzo on Thursday December 15, 2022 @03:42PM
(#34005443)
> Is the to AI going understand that and conduct meetings to hash out the missing details and corrections?
Why not? ChatGPT is nearly there right now.
Re:DeepMind’s AlphaCode Conquers Coding, Performing as Well ...
by nsxwolf on Thursday December 15, 2022 @04:46PM
(#34006242)
Can ChatGPT have suspicions that a requirement isn’t quite right?
Re:DeepMind’s AlphaCode Conquers Coding, Performing as Well ...
by acuozzo on Friday December 16, 2022 @10:18AM
(#34015474)
No, but I'd bet that it'll improve to the point of being able to determine ambiguous requirements and/or ones with insufficient detail.
Given the popularity of "NoCode" I reckon that managers & potential clients are more willing to meet in the middle than one would otherwise assume.
Since programming problems are recursively decomposable into sub-problems, all that would be required is to iterate on the ChatGPT-assisted requirements list until it's unambiguous and then repeat the process for each subproblem until the capabilities of AlphaCode are reached.