Deepseek - China’s New AI Breakthrough

[Raze]

This has collapsed western AI ‘s bubble, as China made a better one with a fraction of the cost and it’s open source.

[Pox]

Just now, Raze said:

This has collapsed western AI ‘s bubble, as China made a better one with a fraction of the cost and it’s open source.

the fraction of the cost thing is a lie  y'all fall for psyop so easily its sad. 

 

[Lyubov]

10 minutes ago, Pox said:

the fraction of the cost thing is a lie  y'all fall for psyop so easily its sad. 

 

I mean let’s wait and see but I wouldn’t be surprised at all if western tech has way overinflated operation / research costs. I said it in here already, this bull run and “to the moon” shit has made that whole sphere ripe with scammers conmen

Edited Monday at 07:20 PM by Lyubov

[AION]

1 hour ago, Pox said:

you're so innocent lmao

@zazenwhy?

[Pox]

15 minutes ago, AION said:

@zazenwhy?

What’s your vision of china as a country? What are the pros and cons of it?

[zazen]

From: https://substack.com/home/post/p-155779348

“The most industry-wobbling thing here is how DeepSeek achieved their feat. 

We can organically GROW intelligence of the general reasoning variety. They clearly demonstrate it. 

DeepSeek-R1-Zero was trained using RL (reinforced learning) directly on a base model without initial SFT (supervised fine tuning) marking a significant departure from conventional methods. This approach allowed the model to explore chain-of-thought (CoT) reasoning and develop capabilities such as self-verification and reflection. 

This suggests that reasoning abilities can be incentivized through RL, rather than relying on supervised data. Please read that again if it didn’t blow your mind.

The models exhibited a self-evolution process where they naturally increased their thinking time and developed complex reasoning behaviors through RL.... An "aha moment" was observed in an intermediate version of DeepSeek-R1-Zero, where the model learned to re-evaluate its initial approach, showcasing the potential of RL to produce unexpected and sophisticated problem-solving strategies.”

“Imagine teaching a child to ride a bike. You could give them a detailed manual (SFT), but they'll likely learn better by trying it themselves (RL), falling, getting up, and gradually improving. DeepSeek-R1's training process is similar - it allows the LLM to "learn by doing" and develop its own reasoning abilities, leading to a more robust and adaptable intelligence.

This approach could revolutionize the field of AI, leading to more capable, efficient, and trustworthy LLMs that can be used for a wider range of applications.

This is the way to AGI.”

Edited Monday at 08:52 PM by zazen

[AION]

@zazen that is insane

[Leo Gura]

But China will not allow its citizens to use this AI for serious independent critical thinking. How will their society handle this, since they are so into censorship?

How can China have serious AI when they don't even allow basic www internet?

Edited Monday at 10:23 PM by Leo Gura

[RightHand]

 

[Thought Art]

 

[hoodrow trillson]

1 hour ago, Leo Gura said:

But China will not allow its citizens to use this AI for serious independent critical thinking. How will their society handle this, since they are so into censorship?

How can China have serious AI when they don't even allow basic www internet?

Same way they implemented TikTok.

[Leo Gura]

But if this AI is so open-source, how they gonna censor it? And what happens to their society when their censorship fails?

[JTL]

15 hours ago, Leo Gura said:

Be careful feeding or uploading any of your data to China.

@Leo Gurawhy though? Is it really any worse than give it to immoral, fascist supporting a american tech companies. Who sell your data till the highest bidder. 

Edited Monday at 10:39 PM by JTL

[AION]

16 minutes ago, Leo Gura said:

But if this AI is so open-source, how they gonna censor it? And what happens to their society when their censorship fails?

AI will help Chinese censorship even more. You are not allowed to ask certain sensitive or political questions to this Chinese machine. I've tried it.

[Thought Art]

Deepseek thinks it’s ChatGPT 

[Pox]

1 hour ago, Leo Gura said:

But China will not allow its citizens to use this AI for serious independent critical thinking. How will their society handle this, since they are so into censorship?

How can China have serious AI when they don't even allow basic www internet?

Exactly🤣 logic is beneath “social stability” in their government’s eyes. It’s ridiculous 

[Pox]

36 minutes ago, JTL said:

@Leo Gurawhy though? Is it really any worse than give it to immoral, fascist supporting an american tech companies. Who sell your data till the highest bidder. 

You don’t know china, just because you hate the west don’t mean china is better

[Lyubov]

1 hour ago, Leo Gura said:

But China will not allow its citizens to use this AI for serious independent critical thinking. How will their society handle this, since they are so into censorship?

How can China have serious AI when they don't even allow basic www internet?

Won’t they just censor certain information the same way Claude or ChatGPT do? For example if you ask it how to build a bomb, extend this to topics China deems sensitive. 

[Leo Gura]

9 minutes ago, Lyubov said:

Won’t they just censor certain information the same way Claude or ChatGPT do? For example if you ask it how to build a bomb, extend this to topics China deems sensitive. 

But China needs so many topics censored I'm not sure that will work. And how do you censor a locally run open-source model?

[hoodrow trillson]

10 minutes ago, Leo Gura said:

But China needs so many topics censored I'm not sure that will work. And how do you censor a locally run open-source model?

You released a gimped version of Deepseek to their own people.  A DeepSeek Lite