How China’s low-cost DeepSeek disrupted Silicon Valley’s AI dominance – Firstpost
DeepSeek could just be the primer in the story with news of several other Chinese artificial intelligence models popping up to give Silicon Valley a jolt
read more
It’s been a couple of days since DeepSeek, a Chinese artificial intelligence (AI) company, rocked the world and global markets, sending American tech titans into a tizzy with its claim that it has built its chatbot at a tiny fraction of the cost and energy-draining data centres that are so popular in the US. Where companies are pouring billions into transcending to the next wave of artificial intelligence.
DeepSeek is everywhere right now on social media and is a burning topic of conversation in every power circle in the world.
So, what do we know now?
DeepSeek was a side project of a Chinese quant hedge fund firm called High-Flyer. Its cost is not just 100 times cheaper but 200 times! It is open-sourced in the true meaning of the term. Many American companies try to solve this problem horizontally by building larger data centres. The Chinese firms are innovating vertically, using new mathematical and engineering approaches.
DeepSeek has now gone viral and is topping the App Store charts, having beaten out the previously undisputed king—ChatGPT.
So how exactly did DeepSeek manage to do this?
Aside from cheaper training, not doing RLHF (Reinforcement Learning From Human Feedback, a machine learning technique that uses human feedback to improve), quantisation, and caching, where is the reduction coming from?
Is this because DeepSeek-R1, a general-purpose AI system, isn’t quantised? Is it subsidised? Or is OpenAI/Anthropic simply charging too much? There are a few basic architectural points compounded together for huge savings.
-
The MoE—Mixture of Experts, a machine learning technique where multiple expert networks or learners are used to break up a problem into homogenous parts.
-
MLA—Multi-Head Latent Attention, probably DeepSeek’s most critical innovation, to make LLMs more efficient.
-
FP8—Floating-point-8-bit, a data format that can be used for training and inference in AI models.
-
Multi-fibre Termination Push-on connectors.
-
Caching, a process that stores multiple copies of data or files in a temporary storage location—or cache—so they can be accessed faster.
-
Cheap electricity
-
Cheaper supplies and costs in general in China.
DeepSeek has also mentioned that it had priced earlier versions to make a small profit. Anthropic and OpenAI were able to charge a premium since they have the best-performing models. Their customers are also primarily Western markets, which are more affluent and can afford to pay more. It is also important to not underestimate China’s goals. Chinese are known to sell products at extremely low prices in order to weaken competitors. We have previously seen them selling products at a loss for 3-5 years in industries such as solar power and electric vehicles until they have the market to themselves and can race ahead technologically.
However, we cannot afford to discredit the fact that DeepSeek has been made at a cheaper rate while using much less electricity. So, what did DeepSeek do that went so right?
-
It optimised smarter by proving that exceptional software can overcome any hardware limitations. Its engineers ensured that they focused on low-level code optimisation to make memory usage efficient. These improvements made sure that performance was not hampered by chip limitations.
-
It trained only the crucial parts by using a technique called Auxiliary Loss Free Load Balancing, which ensured that only the most relevant parts of the model were active and updated. Conventional training of AI models usually involves updating every part, including the parts that don’t have much contribution. This leads to a huge waste of resources. This led to a 95 per cent reduction in GPU usage as compared to other tech giant companies such as Meta.
-
DeepSeek used an innovative technique called Low Rank Key Value (KV) Joint Compression to overcome the challenge of inference when it comes to running AI models, which is highly memory intensive and extremely costly. The KV cache stores key-value pairs that are essential for attention mechanisms, which use up a lot of memory. DeepSeek has found a solution to compressing these key-value pairs, using much less memory storage.
-
And now we circle back to the most important component, DeepSeek’s R1. With R1, DeepSeek basically cracked one of the holy grails of AI, which is getting models to reason step-by-step without relying on mammoth supervised datasets. The DeepSeek-R1-Zero experiment showed the world something extraordinary. Using pure reinforcement learning with carefully crafted reward functions, DeepSeek managed to get models to develop sophisticated reasoning capabilities completely autonomously. This wasn’t purely for troubleshooting or problem-solving; instead, the model organically learnt to generate long chains of thought, self-verify its work, and allocate more computation problems to tougher problems.
Is this a technology fluke? Nope. In fact, DeepSeek could just be the primer in this story with news of several other Chinese AI models popping up to give Silicon Valley a jolt. Minimax and Qwen, both backed by Alibaba and Tencent, are some of the high-profile names that are promising big changes in the AI world. The word on the street is: America built and keeps building bigger and bigger air balloons while China just built an aeroplane!
The author is a freelance journalist and features writer based out of Delhi. Her main areas of focus are politics, social issues, climate change and lifestyle-related topics. Views expressed in the above piece are personal and solely those of the author. They do not necessarily reflect Firstpost’s views.
Post Comment