A Chinese-made artificial intelligence (AI) model called DeepSeek has shot to the top of Apple Store's downloads. Released on 20 January, it quickly impressed AI experts before it got the attention of the entire tech industry - and the world. US President Donald Trump said it was a "wake-up call" for US companies who must focus on "competing to win".
What makes DeepSeek so special is the company's claim that it was built at a fraction of the cost of industry-leading models like OpenAI - because it uses fewer advanced chips. That possibility caused chip-making giant Nvidia to shed almost $600bn (£482bn) of its market value - the biggest one-day loss in US history.
Source: Wikipedia
AI can, at times, make a computer seem like a person. A machine uses the technology to learn and solve problems, typically by being trained on massive amounts of information and recognising patterns. The result is software that can have conversations like a person or predict people's shopping habits. In recent years, it has become best known as the tech behind chatbots such as ChatGPT - and DeepSeek - also known as generative AI. These programs again learn from huge swathes of data.
Millions of people use tools such as ChatGPT to help them with everyday tasks like writing emails, summarising text, and answering questions - and others even use them to help with basic coding and studying.
DeepSeek is the free AI-powered chatbot, which looks, feels and works very much like ChatGPT. That means it's used for many of the same tasks, though exactly how well it works compared to its rivals is up for debate. It is reportedly as powerful as OpenAI's o1 model - released at the end of 2023 - in tasks including mathematics and coding.
Like o1, R1 is a "reasoning" model. These models produce responses incrementally, simulating a process similar to how humans’ reason through problems or ideas. It uses less memory than its rivals, ultimately reducing the cost to perform tasks.
Like many other Chinese AI models, DeepSeek is trained to avoid politically sensitive questions. When the BBC asked the app what happened at Tiananmen Square on 4 June 1989, DeepSeek did not give any details about the topic in China.
But DeepSeek's base model appears to have been trained via accurate sources while introducing a layer of censorship or withholding certain information via an additional safeguarding layer. Deepseek says it has been able to do this cheaply - researchers behind it claim it cost $6m (£4.8m) to train, a fraction of the "over $100m" alluded to by OpenAI boss Sam Altman when discussing GPT-4. DeepSeek's founder reportedly built up a store of Nvidia A100 chips, which have been banned from export to China since September 2022. Some experts believe this collection - which some estimates put at 50,000 - led him to build such a powerful AI model, by pairing these chips with cheaper, less sophisticated ones.
DeepSeek was founded in December 2023 by Liang Wenfeng. Not much is known about Liang, who graduated from Zhejiang University with degrees in electronic information engineering and computer science. But he now finds himself in the international spotlight. Unlike many American AI entrepreneurs who are from Silicon Valley, Mr Liang also has a background in finance. He is the CEO of a hedge fund called High-Flyer, which uses AI to analyse financial data to make investment decisons - what is called quantitative trading. In 2019 High-Flyer became the first quant hedge fund in China to raise over 100 billion yuan ($13m).
DeepSeek's achievements undercut the belief that bigger budgets and top-tier chips are the only ways of advancing AI, a prospect which has created uncertainty about the future of high-performance chips.
My own experience with DeepSeek is that it helps draft responses to emails, improves productivity and reduces cost. I will keep to DeepSeek for now. It is a good alternative to Claude or CoPilot.
References:
DeepSeek: The Chinese AI app that has the world talking, Kelly Ng, Brandon Drenon, Tom Gerken and Marc Cieslak, BBC News, 29 January 2025
Alibaba releases AI model it says surpasses DeepSeek, Eduardo Baptista, Reuters, 30 January 2025
No comments:
Post a Comment