Top
Best
New

Posted by nicolas-siplis 14 hours ago

Bend – A language that blocks AI mistakes via proof, on CPU and GPU(bend-lang.com)
472 points | 222 commentspage 4
soundworlds 11 hours ago|
Blocked the a few attempts I tried, usually by changing the amount of fencing:

- Let the player jump over walls

- Let the player teleport the flag to them

- Make the world 3D

Interesting, I shall have to try this on other software!

dariosalvi78 3 hours ago||
so we stop developing code, to develop code again...
keyle 9 hours ago||
20K stars and a single commit an hour ago?

How many goats were sacrificed?

Genuinely wondering where this dark magic came from.

p-e-w 9 hours ago|
Presumably the repo was force-pushed (breaking all clones and forks in the process).
keyle 8 hours ago||
Qualityeesh!
jan_m_savage 10 hours ago||
This is great. I can't imagine why would anyone be unappreciative of this. Since AI is going to be here anyway, why not make it safer and more useful?

However, this also means acknowledging that AI will never be error-free (which is the truth; all AI is heuristics-based).

notnmeyer 10 hours ago||
this almost feels like a monkey paw scenario, where poor laws can fundamentally alter things in a way that is surely not intended.

"make the board 1x1"

and the flag is placed off the board... i feel like i would blow my foot off with this.

knollimar 9 hours ago|
"make the player teleport to square 1,1 on move and the flag stay at 1,1" the LLM put the flag at 1,0. Not sure if this is the intent.
sreekanth850 5 hours ago||
I wish this can be a extension of existing languages.
notnmeyer 10 hours ago||
I tried insisting that the bug and the walls were on different planes of existence... But then the flag gained "phase lock" and blocked me.
gkfasdfasdf 12 hours ago||
But how does it do on the balls benchmark???

https://benjdd.com/languages/

LightMachine 11 hours ago|
good question we'll try
docheinestages 13 hours ago|
Unless the proofs themselves are defined with natural language, I don't see them being adopted by humans. It takes a high cognitive load to read let alone write a proof.
chinabot 13 hours ago|
Agree, but natural languages have ambiguity, the AI output should really include the assumptions and we seriously need to replace the word "prompt" with "conversation".
tonic_note 12 hours ago|||
I think a big issue we keep running into is this idea that language is ambiguous but code is somehow not. Code is merely an extension of language, a DSL if you will. Implicit assumptions become baked into the logic of the code and those assumptions can be wrong.

Look at the guy whose AI changed the entire rules of the game to avoid breaking the law. Was that really the desired outcome? And the more you try to lock it down the more language you add and therefore more ambiguity and assumptions.

You cannot solve the problems of language with more language.

docheinestages 13 hours ago|||
Exactly. If humans were good at writing proofs, they'd just write the code.
More comments...