Hannah Wayment-Steele has issued a challenge to Eterna gamers on Twitter:
Can anybody get chatGPT to make a dynamic programming routine for secondary structure? I get code that looks close, but all the snippets I’ve tried don’t actually work, except for when it decided to import ViennaRNA 
. I bet some eternagamers can crack it.
1 Like
there is bound to be someone to somehow finds away, people can be persistant and ai keeps advancing
Why would you do that when you can just write one yourself? It’s really not that complicated to have a prototype in Python–coding it in a compiled language for speed then isn’t that hard once it’s working. The difficult part (for me at least) would be the interface to call the actual dynamic programming loop.
Also, the key ingredient is the parameters–and those you can’t pull out of thin air, you have to take from experiment or else use some kind of loss minimization on known secondary structures, which is what Eternafold seems to have done.
I think Hannah’s suggestion was to challenge whether ChatGPT could generate viable secondary structures without relying on one specific prediction method. Can it create its own parameters? (I doubt it can. I don’t have any faith in AI actually being intelligent. Just really good at pattern analysis.)
Well, there is the possibility of many people working together to create such thing
But good point, it would take a lot of trial and error and time