l'empathie mécanique @dpwiz@qoto.org
- Monad m
- m a -> (a -> m b) -> m b
- Previously
- https://mastodon.social/@dpwiz
- Making games
- https://haskell-game.dev
- Games made
- https://icrbow.itch.io
Toots as he pleases.
Joined Apr 2022
@suppi@fosstodon.org @reidrac@social.sdf.org On a more serious note, this very important indeed.
Experts are at least in part formed by the existing deficiencies in the ecosystem. Newcomer's wtfs should be noted and addressed for gradual raising of the language usability waterline.
With games being CrossFit of programming, touching lots of aspects, we have an opportunity to make lots of things better, fast.
@suppi@fosstodon.org @reidrac@social.sdf.org It *is* exciting that it is possible to produce complete games, playable and fun, with mundane concepts and tools.
@rms80 @flaviusb @dpiponi Does it matter, though? There's no "fake addition" that brings you the same results in arithmetic, but without the "real math" behind it. The addition that works is just... addition. Ditto for every other task.
If the next token batch is merely predicted to be some bytes from Metasploit payload your system would be pwned.
@dpiponi It doesn't stop with association though. For some tasks it has pretty strong causal model. I.e. you can test interventions and counterfactuals on it.
@reidrac@social.sdf.org Sweet! Looking forward to adding you game to haskell-game.dev hall of fame 
this is one of the most important launches since lunar race
@
no pressure
@exkclamation @carlos_lunamota Same. And if I decide to wear some, it would be one of those.
@terrorjack and bolts?
@knittingknots2 When you're thinking of children for too much...
Predicting the next token, my stochastic ass...
https://socket.dev/blog/introducing-socket-ai-chatgpt-powered-threat-analysis
The next token is: PWND
@me I used wide + tall too. It's okay, I just think a more rounded approach would work better. Too bad you just can't buy anything less squishy without selling your kidney. It's somehow specialized medical equipment now.
@DrewKadel @dataKnightmare@octodon.social More like they have done their reading... But don't have time to reason properly and systematically and are forced to BS their way through a seminar.
The quality of answers improves significantly when you give them that time ("lets think step by step" etc.).
Given more compute LLMs would be able to do Fermi Estimates on par with the dude himself, if not better. There's nothing putting a hard limit on that, since even the humans are able to do it when motivated.
And with plugins like Wolfram they could build a proper model and perform exact calculations on it.
@dataKnightmare@octodon.social @DrewKadel > LLMs by design are incapable of associating a likelihood value to their output.
> Their output is totally randomly
Sorry, but this is just false. The probability space of "totally random" output is unimaginably huge and most of it not just false, but complete gibberish. Throw a 26-sided dice a few times and compare that totally random result to the GPT output.
To navigate it *at all* requires calculating likelihood and picking only up the sensible stuff.
So "associating a likelihood value to their output" is exactly like the thing works.
@me Wide gets side screens too far. Tall wastes space or forces eye to track objects in 2d instead of a line.
- Monad m
- m a -> (a -> m b) -> m b
- Previously
- https://mastodon.social/@dpwiz
- Making games
- https://haskell-game.dev
- Games made
- https://icrbow.itch.io
Toots as he pleases.
Joined Apr 2022
