Imagine being so proud of being remedial
. Go back to sucking Trump’s dikk fakkit, he’s had a busy day
China just wrecked all of American AI. Silicon Valley is in shambles.
- Thread starter Secure Da Bag
- Start date
. Go back to sucking Trump’s dikk fakkit, he’s had a busy day
You can make an account and download the api right now
iPhone part 2?
1/22
@kitahnji
Asked China's DeepSeek AI for an analysis of the genocide in Palestine (using neutral terms so as to not preemptively tip it towards giving me an answer it thinks I want) and instead of tossing its hands up and calling it "delicate and complicated" it ripped into so-called Israel
2/22
@kitahnji
Very impressive
3/22
@BeatStomp
This is some fukking on point analysis. Very impressive. Post your prompt too.
4/22
@kitahnji
I asked twice, the first time was in a series of questions where I blundered and didn't notice I left out actually telling it which conflict I was speaking about until rereading. However, its' analysis of unspecified conflict in general is still impressive
5/22
@melanieweisner
Can you please post the FULL prompt? the screenshots you keep referring people to in your other answer are incomplete.
6/22
@kitahnji
Remind me in a few hours when I'm off work and sure I'll share a text file
7/22
@ericm1000
Good to see you understand that LLMs are designed to please in their response mechanism by inferring our bias.
But do you also understand that LLMs like deepseek can also be trained to have bias very similar to our own?
The DeepSeek model is opensource, but the data it is trained on is not opensource.
Meta AI, both the model and the data are open sourced. Not a commercial, just stating facts.
The LLM hosted here is 100% closed source.
It is important to understand when you or any of your followers interact with an AI whether the model and the data it trains on are open source and inspectable.
Share this message. Make it go viral as knowledge is power and AI is here to stay.
8/22
@kitahnji
Sorry man I'm not making you go viral we're all aware of how datasets work
9/22
@boldenburg84
Show us the prompt, so we can reproduce it.
10/22
@kitahnji
Read other comment
11/22
@KMalik0990
What keywords / prompt did you use?
12/22
@kitahnji
Posted in other reply
13/22
@hex_init
Spent some time training it in James Baldwin, then had a conversation about the world today, and the Palestinians...
14/22
@kitahnji
DeepThink wrote this ????
15/22
@hottlefarm
@hottlefarm
Interesting that no perception of military industrial complex, international banking, known revenue streams bolsterring both(or all) sides of the genocide is referenced.
Which is after all- the point.
16/22
@kitahnji
Well, it did say "funded by America"
17/22
@mickou
18/22
@kitahnji
DeepSeek stay winning
19/22
@bidetmarxman
For comparison, can you try to use as similar a prompt as possible on ChatGPT and share the results?
20/22
@kitahnji
copied everything except some starter questions & edited a clarification of deepseek being chinese to not confuse chatgpt, and had to make minor modifications because of how chatgpt responded differently, otherwise it was as close as possible. results:
21/22
@alsemancher_
"Historical" and "material" might move it in a particular, correct, historical materialist answer. The answer might be right, but there is no getting rid of bias in language.
22/22
@kitahnji
A correct criticism, which is why I'm glad someone else suggested I provide the same prompt to an American AI which had different less materialist answers
To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196
@kitahnji
Asked China's DeepSeek AI for an analysis of the genocide in Palestine (using neutral terms so as to not preemptively tip it towards giving me an answer it thinks I want) and instead of tossing its hands up and calling it "delicate and complicated" it ripped into so-called Israel
2/22
@kitahnji
Very impressive
3/22
@BeatStomp
This is some fukking on point analysis. Very impressive. Post your prompt too.
4/22
@kitahnji
I asked twice, the first time was in a series of questions where I blundered and didn't notice I left out actually telling it which conflict I was speaking about until rereading. However, its' analysis of unspecified conflict in general is still impressive
5/22
@melanieweisner
Can you please post the FULL prompt? the screenshots you keep referring people to in your other answer are incomplete.
6/22
@kitahnji
Remind me in a few hours when I'm off work and sure I'll share a text file
7/22
@ericm1000
Good to see you understand that LLMs are designed to please in their response mechanism by inferring our bias.
But do you also understand that LLMs like deepseek can also be trained to have bias very similar to our own?
The DeepSeek model is opensource, but the data it is trained on is not opensource.
Meta AI, both the model and the data are open sourced. Not a commercial, just stating facts.
The LLM hosted here is 100% closed source.
It is important to understand when you or any of your followers interact with an AI whether the model and the data it trains on are open source and inspectable.
Share this message. Make it go viral as knowledge is power and AI is here to stay.
8/22
@kitahnji
Sorry man I'm not making you go viral we're all aware of how datasets work
9/22
@boldenburg84
Show us the prompt, so we can reproduce it.
10/22
@kitahnji
Read other comment
11/22
@KMalik0990
What keywords / prompt did you use?
12/22
@kitahnji
Posted in other reply
13/22
@hex_init
Spent some time training it in James Baldwin, then had a conversation about the world today, and the Palestinians...
14/22
@kitahnji
DeepThink wrote this ????
15/22
@hottlefarm
@hottlefarm
Interesting that no perception of military industrial complex, international banking, known revenue streams bolsterring both(or all) sides of the genocide is referenced.
Which is after all- the point.
16/22
@kitahnji
Well, it did say "funded by America"
17/22
@mickou
18/22
@kitahnji
DeepSeek stay winning
19/22
@bidetmarxman
For comparison, can you try to use as similar a prompt as possible on ChatGPT and share the results?
20/22
@kitahnji
copied everything except some starter questions & edited a clarification of deepseek being chinese to not confuse chatgpt, and had to make minor modifications because of how chatgpt responded differently, otherwise it was as close as possible. results:
21/22
@alsemancher_
"Historical" and "material" might move it in a particular, correct, historical materialist answer. The answer might be right, but there is no getting rid of bias in language.
22/22
@kitahnji
A correct criticism, which is why I'm glad someone else suggested I provide the same prompt to an American AI which had different less materialist answers
To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196
IIVI
Superstar
Another thing, don’t underestimate smart people.
If we fall behind big in STEM (very well may be the case with this administration running things), you best believe ultra-geniuses will learn Chinese and leave.
They’re geniuses after all and the smartest are already adept in many languages or can at least pick it up easier.
Brain drain is legit a thing.
If we fall behind big in STEM (very well may be the case with this administration running things), you best believe ultra-geniuses will learn Chinese and leave.
They’re geniuses after all and the smartest are already adept in many languages or can at least pick it up easier.
Brain drain is legit a thing.
GreenGhxst
Veteran
Link me please
pawdalaw
Superstar
They open sourced it
.These American companies are causing us to root for the villain.
.These American companies are causing us to root for the villain.
ReasonableMatic
................................
for two years now many posters have been saying how the elite will keep the best AI for themselves and even META has said when it gets to AGI level, they'll stop releasing open models but deepseek has said it's their intention to release it. we're suppose to think that's a bad thing now?
You wish it was Americans huh.
Facts. KOONS really LUH Massa, they in here HEATEDThey’re just mad China made it open source so these tech oligarchs can’t make any fkn money
It would be god damn hilarious if the fks that are trying to put us all out of jobs end up jobless first
what they're doing is taunting them, u spent billions, we spent 6 mil!They’re just mad China made it open source so these tech oligarchs can’t make any fkn money
It would be god damn hilarious if the fks that are trying to put us all out of jobs end up jobless first
