I just love how we've now spent far more computational resources on generating funny pictures and spitting out flawed code than was ever spent on running formal verfication and exhaustive testing on important software because it was thought to be "too expensive".

@richcarl
This gave me pause for thought.

I think it's because machine learning is relatively easy - it's just a technical problem, and we're good at solving those.

Changing development practices, languages and tools is a social and cultural change - a people problem - and that is very hard.

@richcarl OMG my two biggest AI enthusiast coworkers were just yesterday saying “we don’t want to write too many Selenium tests they’re really resource intensive” 😂😂
Amazing how a zillion dollar company buying market share can distort your perception of sustainability.
@richcarl One looks fascinating to people with money and no understanding of the technology, the other doesn't. Somewhat sarcastic, but I think it's likely close to the truth.
@richcarl Yeah, but which one gives the public bigger dopamine surges?
@richcarl the dull part of it is that there is a lot of effort required to do formal verification amd exhaustive testing while instructing an LLM and hoping for the best is trivial effort.
@richcarl @lawik GIGO applies though
@mirabilos @richcarl well companies have a poor record of separating well-articulated garbage from valuable insights...
@richcarl No one gets rich testing code.

@richcarl

It's called priorities.

@richcarl we expend resources on things to further conflict, prestige or money. People can see the money behind AI cats. I'm afraid the only thing that would drive forward formal verification is conflict.
@richcarl
Not to mention fake money...  
@richcarl How do I monetize tested, correct software?
(/s)
@richcarl I guess nobody sold formal verification as “it allows you to save money by firing developers and replacing them with X”, or “it’s so easy you yourself can do it”.
@richcarl okay that’s because formal verification stops you from making mistakes while AI helps you make them faster
@richcarl
"too expensive" === "too less fun"
@richcarl was there ever a better opportunity to say « the sarcasm is strong with this one »?
@richcarl I really feel like llm making formal methods proofs will be an incredible moment