China's new AI Model: DeepSeek

[Amaterasu]

Folks online tried the model and noted that replies are moderated and curated heavily.

Kerna choked when counting to 15

 

[GenXYZ]

$NVDA is the Drug supplier leh.
All the AI stuff using their cards.
AI = Drugs
NIVIDA = supplier
Chatgpt, Deepseek, Llama, etc= Different Drug brand.
Someone trying to sell the News and profit later.

with Deepseek people are theorizing that they might not need the latest and greatest nvidia cards and in the numbers that they think they need. so, nvda growth might be affected, considering it's price.

 

[Suporter]

Sounds kinky...

 

[_zeph_]

Folks online tried the model and noted that replies are moderated and curated heavily.

Kerna choked when counting to 15

but chatGPT also has built-in woke bs in it too. but lucky for me, my work involves coding and troubleshooting. nothing to do with politics so, it should be fine for me and most of the people using it.

 

[jeffprobst]

Easier for them to scam?

 

[evilrion]

hokays noted

 

[econ food]

I just asked a few question on it, very impressive.

Got ask this? Toto sg next draw?

 

[NoSouL]

not bad for the 1st 2 sec, the answer is quite ok.

but it will auto censored it after like 1 or 2 sec.

you will need to screenshot it fast before they censor it.

 

[tExtra]

not bad for the 1st 2 sec, the answer is quite ok.

but then it will auto censored it after that.

Ya. It’s okay for coding. But recent history and current affairs, especially their own all BKY..

 

[Amaterasu]

but chatGPT also has built-in woke bs in it too. but lucky for me, my work involves coding and troubleshooting. nothing to do with politics so, it should be fine for me and most of the people using it.

Ya. It’s okay for coding. But recent history and current affairs, especially their own all BKY..

Yes but not everyone use it solely for development. The model is also applicable to general day to day task for the layman. Some people even use it for fact checking too.

And OpenAI is not the only AI model around, you have got Perplexity etc whereas in China you got only 1 choice.

 

[Humster]

with Deepseek people are theorizing that they might not need the latest and greatest nvidia cards and in the numbers that they think they need. so, nvda growth might be affected, considering it's price.

You seriously don't need the greatest and most expsensive Nvidia cards to do stuff, even currently.
The main ppl that uses those 1337 cards are Giant corps, google, MS, Huawei, Amazon, softbank, etc.
Average Joe users i.e home users or small home base business don't even need those.
Heck the latest Intel 258V or Ryzen 365H can liao.

 

[GenXYZ]

You seriously don't need the greatest and most expsensive Nvidia cards to do stuff, even currently.
The main ppl that uses those 1337 cards are Giant corps, google, MS, Huawei, Amazon, softbank, etc.
Average Joe users i.e home users or small home base business don't even need those.
Heck the latest Intel 258V or Ryzen 365H can liao.

I’m referring to corps like those involved in the 500b AI Infra deal. When i say ‘people’ I mean investors/analysts. My point still stands, DeepSeek was able to achieve close to what OpenAI has achieved in the benchmarks with way fewer resources. Market is reacting to the fact that the demand for nvda’s v/a/h100 chips might go down.

 

[Darkorochix]

Folks online tried the model and noted that replies are moderated and curated heavily.

Kerna choked when counting to 15

You download it to your PC and train it yourself.
Can even make it say “XJP is Winnie and like to eat covid meat at Tiananmen on a tank.”

Every apps in China will be bound to China Censorship Law.

 

[Amaterasu]

You download it to your PC and train it yourself.
Can even make it say “XJP is Winnie and like to eat covid meat at Tiananmen on a tank.”

Every apps in China will be bound to China Censorship Law.

I'll leave it to the open source community to first scrutinize how tethered (the model) is it to the chinese government before even consider putting it on my machine.

 

[askxyz]

https://futurism.com/deepseek-ai-answer-tiananmen-square-massacre

 

[Shion]

imo this is why deepseek will never be that successful despite the hype

 

[artncraft]

imo this is why deepseek will never be that successful despite the hype

All the AIs are biased and censored. It is only how well they hide them.

 

[nasfieldjohn]

https://futurism.com/deepseek-ai-answer-tiananmen-square-massacre

Use keyword in chatgpt and other ai chat works though

 

[FireEmblem]

Oh no I am going to get banned, faster delete.