How a low -cost AI model of Deepseek violates the market | Boom – Boom

Almost $ 1 trillion ($ 600 billion) was deleted by the value of the NVIDIA artificial intelligence manufacturer overnight on Monday, when a little -known Chinese start -up, Deepeek, threatened to increase the US technology market.

While NVIDIA has suffered the largest one-day loss in the history of ShareMarket, other technological giants-Microsoft, Alphabet and Amazon, which invest seriously in competitive AI tools, including Chatgpt and Gemini, were also hit.

The route is caused by the shock of investors by the claimed presentation of the new R1 chatbot on Deepseek. The Chinese AI has been reported to be more advanced than its competitors and more cheaply to develop.

The Depepeek R1 is rising, becoming the best free application on Apple’s App Store, as US technology and related stock prices have dropped dramatically.

Why did technological stocks dive deeply

The market was surprised by Deepseek, providing what more technology is, but comparable efficiency.

This drastically changed the expectations of the market for the computing power, showing more can be done in less. It also compromises the competitiveness of existing AI products and developments of US technology.

Shares prices are guided by market expectations. The requested implementation of Deepseek R1 has sparked a major revision of expectations of what is technologically possible and how cheap AI can develop and operate.

Investors quickly included the news of a cheap Chinese competitor in shares prices, expecting that this new participant can break the market and erode the competitive advantage of existing leaders.

Who is Deepseek and what is R1?

Deepseek was founded in 2023 by the Chinese High Flyer Hedge Fund, which has been exclusively using AI in the 2021 trade.

Deepseek develops large language models (LLM) that can be the basis of chatbots and other AI -based tools. The R1 is the most recent iteration of the Deepseek chatbot and the main model. It is based on larger versions of generative models of AI developed by Deepseek and significant amounts of data, but is a surprising jump forward in performance and costs.

Investors in technology believe that R1 matches or outperform competitors, including Chatgpt 4.O1 of Openai on numerous indicators.

However, there are some key differences:

1. The model, which is the basis of R1, works in a much less intense way. It is much more cheaper to develop and execute, requiring less data and computing power.

2. The training of the model was possible, despite the prohibition of exports to the United States, preventing Chinese companies such as Deepseek from having access to chips from US companies like NVIDIA. The Biden administration has introduced laws restricting the sale of certain computer chips and machines in China in progress designed to block its opponent from accessing some of the most modern technologies in the world.

3. Training and data uploaded to R1 are found on servers in China. Given concerns about data confidentiality and intellectual property, they have already been raised for US-based companies, with data under the jurisdiction of the Chinese Communist Party (CCP) being even more concerned.

4. The Chatbot program code is free for download, reading and changing, as opposed to Chatgpt. However, this is somewhat false transparency – the important model is the important model, not the chatbot code.

5. It is known that R1 censors its answers in accordance with the values ​​of the Chinese Communist Party.

The future of AI and technological stocks

It is not known whether this collapse in the price of technological stocks is an irrational panic that will turn or simply reflect the right prices. Future costs and benefits of AI are still uncertain.

This is both a technological and economic issue.

Technologically, it is yet to be seen whether R1 really requires less computing power and less data on training and use.

Economically there are potential winners and losers. AI users can earn more expensive access to AI, and LLMS more special, leading to increased acceptance and related profit of productivity. Existing manufacturers such as NVIDIA may lose what was a market with a little real competitors.

More broadly, society can benefit from less computational intensive and therefore more energetically efficiently, AI. However, the geopolitical risk of an individual country capturing the market, along with concerns about data privacy, intellectual property and censorship, may exceed the benefits.

Michael J. Donne, Professor of Accounting and Business Information Systems, The University of Melbourne and Matt Pinkuk, a professor of financial accounting, The University of Melbourne

This article was reissued by the Creative Commons license. Read the original article.