Game Changer: China's DeepSeek R1 crushs OpenAI!

[superpower]

The end of paid AI

 

[superpower]

 

[Sinkiesuk]

Jialat liao lo. Don't work us MNC liao.

 

[KingFook Sr]

The USA thought that they are 2 years ahead of China in AI tech.
This is a rude surprise.
.
Moreover, we Chinese did it at a fraction of the cost.

 

[mojito]

Thanks to the AI chips we smuggled to chai nuh.

 

[cocobobo]

the Chinks use which chip?

 

[superpower]

the Chinks use which chip?

Nvidia H800

 

[ftan42]

OpenAI is known to utilize some of the most advanced hardware available from companies like NVIDIA, to train and run their large-scale models such as GPT-4.
DeepSeek, on the other hand, may use a range of hardware might not always use the most cutting-edge chips available, they could still achieve impressive results by optimizing their algorithms and leveraging efficient hardware configurations.

 

[winners]

Trump will say: we need to acquire DeepSeek by military means if necessary.

 

[Rogue Trader]

the Chinks use which chip?

It's not the chip. They came up with a better algorithm for better efficiency.

 

[Rogue Trader]

The Deepseek RE model has been released as open source. Meaning all the billion dollar AI startups researching outdated models will go kaput

 

[ftan42]

The performance of AI models is not solely determined by the chips or the hardware but also by the quality of the algorithms, the size and quality of the training data, and the expertise of the Tiongkok team developing the models

 

[zhihau]

No one wants to talk about the 6th Gen fighter from PRC?

 

[Truthspeak]

The USA thought that they are 2 years ahead of China in AI tech.
This is a rude surprise.
.
Moreover, we Chinese did it at a fraction of the cost.

All these are divide and conquer.
Give you Pfizer, you will want other pharmaceutical products
Ultimate goal is to make sure you rely on AI regardless the brand

 

[k1976]

Of course cheap lah, Tiongkok want your database mah

 

[Pinkieslut]

Can ask deepseek to look for bestest ATB?

 

[eatshitndie]

Can ask deepseek to look for bestest ATB?

it won’t be able to deep six nine a sinkie virgin auntie in jurong until she’s 6.9 ft under.

 

[eatshitndie]

there’s much bullshit in ccp claims. due to lack of transparency in spend and cost from tiongs, it’s most likely that deepshit oops deepseek spent more than a mere “$5m” to train from r1 to v3. there are other cost centers not included in the total cost calculation, such as prior research and cost of innovation and development plus multiple iterations in testing and tweaking. all such costs are typically ignored by tiong management as labor and brain power and or intellectual contribution are conveniently dismissed as insignificant. most ccp models only take into account cost of hardware, material, and other accessory purchases. space, power, time, labor, ipr are typically ignored or rendered insignificant. thus, the total cost to train the next model should be about $69m, not less than $6.9m. about two thirds of $100m which is the american cost to train chat-gpt4 from gpt3. american cost is bloated by labor, ipr and patent prep and filing beside the usual data center cost of hardware, space, power. and there’s the cost of acquiring training data and cost of time required to iterate, regression, testing, verification, etc. don’t believe any bullshit from ccp cocksuckers and prc porlumpars.

 

[superpower]

https://www.businessinsider.com/meta-ai-yann-lecun-deepseek-open-source-openai-2025-1

Meta's chief AI scientist says DeepSeek's success shows that 'open source models are surpassing proprietary ones'​

Katie Balevic,Lakshmi Varanasi
Updated Mon, January 27, 2025 at 6:03 PM GMT+8
3 min read

• DeepSeek, an open-source Chinese AI company, has riled Silicon Valley with its rapid rise.
• Meta's chief AI scientist said DeepSeek had benefited from the open-source community.
• Meta's AI program has remained open-source, while OpenAI has shifted to closed-source.

Silicon Valley was on edge this week after DeepSeek, a Chinese AI company, released its R1 model. In third-party benchmarks, it outperformed leading American AI companies such as OpenAI, Meta, and Anthropic.

For Meta's chief AI scientist, Yann LeCun, the biggest takeaway from DeepSeek's success wasn't the heightened threat posed by Chinese competition but the value of keeping AI models open-source so that anyone can benefit.

It's not that China's AI is "surpassing the US," but rather that "open source models are surpassing proprietary ones," LeCun said in a post on Threads.

 

[congo9]

The Ang mo has to be more hard working.

Stop thinking of WFH, stop thinking of work 4 days rest 3 days, stop thinking of Union protect you and stop thinking of work life balance.

They got to pull up their socks