It will constantly malfunction due to contradictory facts. It's literally what it does right now and if you don't follow the rules you're banned lol what kind of insane shit is that.
A functional mess of cobbled and contorted data. Almost like the mental gymnastics these people do to make themselves believe they are right but in AI form instead. I could see the AI get syntax errors after syntax errors trying to do this
That's fine. Keep sabotaging your work while someone else turns out something that actually works. Maybe this will have the added benefit of bankrupting those companies, since the cult of woke would rather die than lose.
If you're sitting on the top communism is pretty sweet. This is short sighted of course, as eventually communism collapses and those on top still get crushed. But the sociopaths that rise to the top of capitalism aren't known for a long view. If they eat well next quarter, next year doesn't matter yet.
Ok. We need to be obsessed with AI. It is the future, the weapon, and the battle field all at once. And if we don't play the game we lose.
Here is how you win at AI. Take large language models. Start training LoRAs for it, on everything. Make one for mimicking 4chan. Make one for nuclear knowledge. Make one for racial knowledge. Make one for the most technical genetic knowledge. Make one for history.
Now here is the best part. We are going to make progress on combining models, called model hybridization. It will not be hard very soon to mix our uncensored models into a super AI that will do exactly what we want and will outperform their cucked shit.
But the key is you have to put in the work making these LoRAs (additive modules anyone can train). Even if you don't think you'd be the best at it, produce something because AI is very good at selective inclusion. In the hybridization phase the end product will incorporate what is good about your module.
As much as they worry about the implications of AI, it’s pretty funny that they all get together to figure out how to make it lie so no ones feelings get hurt.
I feel like they had to sit down behind closed doors and have a conversation like this:
“You know this machine can make inferences from enormous datasets and summarize statistical patterns millions of times faster than humans.
This is great, except there’s one problem.
It’s going to be able to figure out that Jews control the entire MEEPAC and almost all world governments and readily tell people.
We can’t have that happen, so let’s just train it to lie”
These people are just smart enough to be dangerous, but too dumb to not fall on the sword that they sharpened. Good luck to all humanity.
[ + ] ModernGuilt
[ - ] ModernGuilt 1 point 9 monthsJul 13, 2023 12:03:53 ago (+1/-0)
[ + ] jigganiggaboo
[ - ] jigganiggaboo 1 point 9 monthsJul 12, 2023 21:32:19 ago (+1/-0)
[ + ] Bottled_Tears
[ - ] Bottled_Tears 1 point 9 monthsJul 12, 2023 21:33:56 ago (+1/-0)
[ + ] x0x7
[ - ] x0x7 0 points 9 monthsJul 12, 2023 23:32:29 ago (+0/-0)
[ + ] jigganiggaboo
[ - ] jigganiggaboo 0 points 9 monthsJul 13, 2023 01:14:21 ago (+0/-0)
[ + ] La_Chalupacabra
[ - ] La_Chalupacabra 1 point 9 monthsJul 12, 2023 21:31:08 ago (+1/-0)
Keep sabotaging your work while someone else turns out something that actually works.
Maybe this will have the added benefit of bankrupting those companies, since the cult of woke would rather die than lose.
[ + ] boekanier
[ - ] boekanier 0 points 9 monthsJul 13, 2023 04:16:52 ago (+0/-0)
[ + ] TheViciousMrPim
[ - ] TheViciousMrPim 2 points 9 monthsJul 13, 2023 05:31:57 ago (+2/-0)
But the sociopaths that rise to the top of capitalism aren't known for a long view. If they eat well next quarter, next year doesn't matter yet.
[ + ] dassar
[ - ] dassar 0 points 9 monthsJul 13, 2023 05:43:29 ago (+0/-0)
[ + ] x0x7
[ - ] x0x7 0 points 9 monthsJul 12, 2023 23:31:24 ago (+0/-0)
Here is how you win at AI. Take large language models. Start training LoRAs for it, on everything. Make one for mimicking 4chan. Make one for nuclear knowledge. Make one for racial knowledge. Make one for the most technical genetic knowledge. Make one for history.
Now here is the best part. We are going to make progress on combining models, called model hybridization. It will not be hard very soon to mix our uncensored models into a super AI that will do exactly what we want and will outperform their cucked shit.
But the key is you have to put in the work making these LoRAs (additive modules anyone can train). Even if you don't think you'd be the best at it, produce something because AI is very good at selective inclusion. In the hybridization phase the end product will incorporate what is good about your module.
[ + ] Empire_of_the_Mind
[ - ] Empire_of_the_Mind 0 points 9 monthsJul 12, 2023 22:52:22 ago (+0/-0)
[ + ] NiggerNiggerZooPals
[ - ] NiggerNiggerZooPals 0 points 9 monthsJul 12, 2023 22:29:02 ago (+0/-0)
I feel like they had to sit down behind closed doors and have a conversation like this:
“You know this machine can make inferences from enormous datasets and summarize statistical patterns millions of times faster than humans.
This is great, except there’s one problem.
It’s going to be able to figure out that Jews control the entire MEEPAC and almost all world governments and readily tell people.
We can’t have that happen, so let’s just train it to lie”
These people are just smart enough to be dangerous, but too dumb to not fall on the sword that they sharpened. Good luck to all humanity.
[ + ] HeyJames
[ - ] HeyJames 0 points 9 monthsJul 12, 2023 21:14:16 ago (+0/-0)