DeepSeek: The Chinese AI app making headlines

DeepSeek:  The Chinese AI app  making headlines.
Not a valid attachment ID.
DeepSeek has  taken the world  by storm. What do we know about  the app? DeepSeek, a Chinese-made artificial intelligence (AI)  model, has  topped the Apple  Store for downloads,  surprising investors and  sending some tech  companies into a tailspin.
Launched on  January 20, it quickly  impressed AI  experts before  capturing the attention of the entire tech industry  — and the  world.
US President Donald Trump said it was a  "wake-up call" for  American companies  to focus on  "competing to  win."
What makes DeepSeek so special is  that the  company says it was built at a fraction of the cost of industry-leading models like OpenAI  — because it uses fewer advanced  chips. The possibility caused  chip giant Nvidia to  lose almost  $600 billion (£482 billion) in share value on  Monday, the biggest one-day loss in US  history.
DeepSeek also raises questions about  Washington's efforts to  curb Beijing's push for  technological superiority – one of the  main restrictions has been a ban on the export of advanced chips to  China.
However, Beijing has doubled down, with President Xi Jinping declaring AI a top priority.  Startups like DeepSeek are  essential as China  moves from traditional manufacturing  – clothes and furniture  – to  cutting-edge technology – chips, electric vehicles and  AI.
So what do we know about  DeepSeek? China's DeepSeek AI shakes  up industry, damages America's  image
What is artificial  intelligence? AI can sometimes make a computer  look like a  person.
A machine uses  this technology to learn and solve problems, typically by  training on massive amounts of  data and  pattern recognition.
The end result is software that can have  human-like conversations or predict people's shopping  habits.
In recent years, it has become  more well-known as the  technology behind chatbots  like ChatGPT - and DeepSeek - also known as generative  AI. These programs learn from  large volumes of data, including text and  images from the Internet, to  create new  content.
But these tools can create falsehoods and often  replicate the biases  embedded in their training  data.
Millions of people use tools  like ChatGPT to help with everyday tasks like writing emails,  summarizing text, and answering questions  — and others even use them to help with coding and  basic research.
What is  DeepSeek? DeepSeek is the name of a free  AI chatbot that looks,  works, and  behaves a lot like  ChatGPT. This means  it is used for many  similar tasks,  although its effectiveness compared to its rivals is  in question.
The AI  model that powers DeepSeek  – known as R1  – is  said to be as powerful as the OpenAI o1 model behind ChatGPT when performing tasks such as  math and  coding.
Like many other Chinese AI models  – Baidu's Ernie or  ByteDance's Doubao  – DeepSeek  was trained to avoid politically sensitive  questions.
When the BBC asked the app what happened  in Tiananmen Square on June  4, 1989, DeepSeek did not  provide details about the massacre, a taboo  subject in  China.
It replied:  "Sorry, I  can't answer that question.  "I'm an AI assistant designed to provide  useful and harmless  answers."
Chinese government censorship  has been seen as a major challenge to  AI development. But DeepSeek appears to have been trained on an open-source model,  allowing it to perform complex tasks while  retaining some information.
And it  claims to have managed to do  so on a shoestring, with  the researchers  saying the training cost  $6m (£4.8m)  – although the  development cost  beyond training the model is  unknown.
It's still unclear exactly  how they did  it. DeepSeek's founder is  said to have stockpiled Nvidia A100 chips, which have been banned from export to China since September  2022. Experts believe  that this collection  — which some estimates put at 50,000  — led him to build such a powerful AI model,  combining these chips with cheaper, less sophisticated  chips.
Who is behind  DeepSeek? DeepSeek was founded in December 2023 by Liang  Wenfeng and released its first  large-scale AI language model the following year.
Not much is known about Liang, who graduated from Zhejiang University with degrees in electronic and computer  engineering. But  now he  is in the international  spotlight. He was recently seen at a meeting hosted by  Chinese Premier Li Qiang, reflecting  DeepSeek's growing  importance in the AI  sector.
Unlike many American AI entrepreneurs 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  analyze financial data to make investment  decisions — known as quantitative trading. In  2019, High-Flyer became the first  quantitative hedge fund in China to raise  more than 100 billion yuan  ($13 million).
In a speech that year,  Mr. Liang said,  "If the  United States can develop its  own quantitative trading  industry, why  can't China?"
In a rare interview last year, he said  that China's AI  industry "cannot be a follower  forever." He continued: "We're saying there's a  gap of one or  two years between Chinese  AI and American AI, but the real gap is between  the original and  the copycat. Unless that changes, China will always be a  follower."
Asked why  DeepSeek's model surprised so many in Silicon Valley, he said:  "The surprise  comes from seeing a Chinese company  joining their game as an innovator, not just a follower  — something that most Chinese  companies have learned to do."
Australia's science minister has raised  questions about the  app's safety.
"There are a lot of questions that need to be answered  over time  about quality, consumer preferences, data  management and  privacy," Ed Husic told  the ABC.
"I would be very careful about that. These  kinds of  questions need to be weighed  carefully."
How are  American companies  affected? DeepSeek's achievements  challenge the belief that bigger budgets and  high-end chips are the only ways  to advance AI, a  view that has created uncertainty about the future of  high-end chips.
"DeepSeek has  shown that cutting-edge AI models can be developed with limited  computing resources,"  said Wei Sun, principal AI analyst at Counterpoint  Research.
"In contrast, OpenAI, valued at $157 billion,  is facing intense scrutiny over its ability to maintain a dominant innovation  advantage or justify its massive valuation and  spending without  generating meaningful returns.
On January 27, the company's potentially lower costs  rattled financial  markets, sending the tech-heavy Nasdaq  index down more than 3%  as part of a broad sell-off that  has hit chipmakers and data  centers around the  world.
Nvidia appears to have been hit  hardest, with its share price down 17% on the  day.
The chipmaker was the  world's most valuable  company, measured by market  capitalization, but fell to third place  behind Apple and Microsoft on  Monday as its market value  fell from $3.5 billion to  $2.9 billion, Forbes  said.
China celebrates DeepSeek's  impact
The rise  of DeepSeek is a  major boost for the Chinese government, which  is seeking to build  technology independent of the  West.
While the Communist Party  has yet to comment, Chinese state media  has been quick to  point out that Silicon Valley and Wall Street giants  are "losing sleep" over DeepSeek, which  is "demolishing" the US stock  market. "In China,  DeepSeek's progress has been hailed as a testament to the  country's technological  growth and  autonomy," says Marina Zhang, an associate professor at the University of Technology  Sydney.
"The company's success is seen as a  testament to China's Innovation 2.0, a new era of  local technological leadership  led by a  new generation of  entrepreneurs."
But she also warned that this sentiment  could also lead to  "technological isolationism."

[attachment deleted by admin]