Close Menu
Karachi Chronicle
  • Home
  • AI
  • Business
  • Entertainment
  • Fashion
  • Politics
  • Sports
  • Tech
  • World

Subscribe to Updates

Subscribe to our newsletter and never miss our latest news

Subscribe my Newsletter for New Posts & tips Let's stay updated!

What's Hot

The world’s largest air force with the F-35 fleet in 2025

AI systems learn from many types of scientific information and run experiments to discover new materials | MIT News

Among the most troublesome relationships in healthcare AI

Facebook X (Twitter) Instagram
  • Home
  • About us
  • Advertise
  • Contact us
  • DMCA
  • Privacy Policy
  • Terms & Conditions
Facebook X (Twitter) Instagram Pinterest Vimeo
Karachi Chronicle
  • Home
  • AI
  • Business
  • Entertainment
  • Fashion
  • Politics
  • Sports
  • Tech
  • World
Karachi Chronicle
You are at:Home » DARIO Amodei will challenge Deepseek’s $ 6 million AI story: Human thinking about the latest AI movements in China
AI

DARIO Amodei will challenge Deepseek’s $ 6 million AI story: Human thinking about the latest AI movements in China

Adnan MaharBy Adnan MaharJanuary 30, 2025No Comments6 Mins Read0 Views
Facebook Twitter Pinterest Telegram LinkedIn Tumblr Email Reddit
Share
Facebook Twitter LinkedIn Pinterest WhatsApp Email


Participate in the new and weekly newsletter for the latest updates and exclusive content on the leading AI coverage in the industry. learn more

The AI ​​world shakes last week when Deepseek, a Chinese AI startup, announced the latest language model Deepseek-R1. The announcement has wiped out nearly $ 200 billion from NVIDIA’s market value, causing a wide range of market sales that caused a fierce debate on the future of AI development.

The story that emerged immediately suggests that DeepSeek has fundamentally confused economics of building advanced AI systems, and maybe a few billions of US companies have spent billions of dollars. Achieved for $ 6 million. This interpretation sent a shock wave by companies such as Openai, Anthropic, and Google through a large -scale investment in a large -scale investment in calculating infrastructure to maintain the advantage of technology.

However, in the turbulent headline and breathtaking headlines, Dario, a co -founder of mankind and one of the pioneering researchers behind today’s large language model (LLMS).・ Amday has published a detailed analysis that provides a more subtle perspective on the results of deep seek. His blog post provides some important insights, surviving hysteria and what DeepSeek has actually achieved and what it means for the future of AI development.

Restoring the understanding of DeepSeek’s presentation, four important insights from Amodei’s analysis are shown below.

1. The story of the “$ 6 million model” misses important context

According to Amodei, DeepSeek’s reported development costs must be seen via a wider lens. He will challenge popular interpretation directly:

“DeepSeek does not cost $ 6 million for billions of AI companies for $ 6 million.” I can only speak humanity, but Claude 3.5 Sonnet is a medium -sized model that costs tens of millions of dollars (accurate numbers). Also, 3.5 Sonnets were not trained in a larger and more expensive or more expensive way (contrary to some rumors). “

This shocking revelation fundamentally changes the story, focusing on DeepSeek’s cost efficiency. Sonnet is trained 9 to 12 months ago and thinks that many tasks are better than Deepseek models, which is not an innovative breakthrough, but a natural progress of AI development costs. It looks more lined up along.

Timing and context are also important. Following the historical trend of AI development cost reduction, which is estimated to be about four times a year, the cost structure of DeepSeek seems to be mainly trends, rather than dramatically proceeding the curves.

2. Deepseek-V3 was the actual technical result instead of R1

Market and media focused on Deepseek’s R1 model, but Amodei pointed out that the company’s more important innovation has come before.

“Deepseek-V3 is actually a true innovation, and it should have been a month ago (certainly so). Several important tasks as a prior model. It seems to be approaching the state -of -the -art US model, but it costs a lot of training.

The distinction between V3 and R1 is important to understand the true technical progress of DeepSeek. V3 represents a real engineering innovation, especially when managing the model “key value cache” and pushing the boundary between the mixed expert (MOE) method.

This insight can help you explain why the dramatic response to the R1 was forgotten. R1 basically adds a enhanced learning function to the foundation of V3. This is the step that multiple companies are currently using the model.

3. Investing in total corporate has revealed a different situation

Perhaps the most obvious aspect of Amodei’s analysis is about the overall investment of Deepseek’s AI development.

“It has been reported -we can’t be convinced that it is true -Deep Shik actually had 50,000 hopper genrey chips. These 50,000 hopper chips are ~ 1 billion. The dollar’s order is different from the US AI lab (which is different from the training of each model).

This revelation dramatically reconstructs the story of DeepSeek resource efficiency. The company may have achieved impressive results through individual model training, but the overall investment in AI development seems to be almost comparable to the American counter part.

The distinction between model training costs and total corporate investment emphasizes the continuous importance of substantial resources in AI development. Although it can improve engineering efficiency, it has been suggested that maintaining AI competitiveness still requires large capital investment.

4. The current “crossover point” is temporary

Amodei explains that the current moment in AI development is unique but someday.

“Therefore, we are in the interesting” crossover point. ” He writes there that some companies can create a legitimate inference model. ” “As everyone moves up the scaling curve of these models further up, this is no longer true.”

This observation result offers important contexts to understand the current status of AI competition. The ability of multiple companies to achieve similar results in the inference function represents a temporary phenomenon, not a new situation.

This is important for the future of AI development. Companies continue to expand their models, especially in resource -aged resources for reinforcement learning, so this field may be re -distinguished based on those who can invest in training and infrastructure. This suggests that DeepSeek has achieved an impressive milestone, but has not fundamentally changed the long -term economics of AI development.

True cost of AI construction: What Amodei analysis reveals

Detailed analysis on Amodei’s Deepseek results reveals the actual economics of reducing market speculation for several weeks and building advanced AI systems. His blog posting systematically dismantling both panic and enthusiasm following Deepseek’s presentation, indicating how the 6 million dollar model training cost is suitable in steady marches of AI development.

The market and the media were drawn to a simple story, and it was attractive that Chinese companies dramatically lower US AI development costs. However, Amodei’s failure reveals more complex reality. Deepsque’s total investment, especially reported $ 1 billion computing hardware, reflects the expenditure of the US counter part.

The moment of cost parity during the United States and China’s AI development marks what Amodei calls “crossover point”. This is a temporary window where multiple companies can achieve similar results. His analysis suggests that this window is closed because the AI ​​function is moving forward and training intensifies. This field can return to an advantageous organization with the deepest resources.

The construction of advanced AI is still an expensive effort, and Amodei’s cautious screening indicates why it is necessary to examine the full range of investment to measure its true cost. It may prove that his systematic decoction of DeepSeek’s results is ultimately more important than the first announcement that caused such a turbulence in the market.

Daily insights on business use case in VB every day

If you want to impress your boss, VB Daily covers it. From regulatory shifts to actual development, we provide internal scoops about what companies are doing in the generated AI, so you can share the largest ROI insight.

Please read the privacy policy

Thank you for subscribing. Please see this VB Newsletter.

An error has occurred.



Source link

Share. Facebook Twitter Pinterest LinkedIn Reddit WhatsApp Telegram Email
Previous ArticleIRGC uses AI for air defense and navy operations
Next Article SoftBank discusses Openai to invest up to $ 25 billion.
Adnan Mahar
  • Website

Adnan is a passionate doctor from Pakistan with a keen interest in exploring the world of politics, sports, and international affairs. As an avid reader and lifelong learner, he is deeply committed to sharing insights, perspectives, and thought-provoking ideas. His journey combines a love for knowledge with an analytical approach to current events, aiming to inspire meaningful conversations and broaden understanding across a wide range of topics.

Related Posts

AI systems learn from many types of scientific information and run experiments to discover new materials | MIT News

September 25, 2025

Among the most troublesome relationships in healthcare AI

September 25, 2025

Does access to AI become a fundamental human right? Sam Altman says, “Everyone would want…”

September 23, 2025
Leave A Reply Cancel Reply

Top Posts

20 Most Anticipated Sex Movies of 2025

January 22, 2025461 Views

President Trump’s SEC nominee Paul Atkins marries multi-billion dollar roof fortune

December 14, 2024122 Views

How to tell the difference between fake and genuine Adidas Sambas

December 26, 202486 Views

Alice Munro’s Passive Voice | New Yorker

December 23, 202474 Views
Don't Miss
AI September 25, 2025

AI systems learn from many types of scientific information and run experiments to discover new materials | MIT News

Machine learning models can speed up discovery of new materials by making predictions and proposing…

Among the most troublesome relationships in healthcare AI

Does access to AI become a fundamental human right? Sam Altman says, “Everyone would want…”

Google’s Gemini AI is on TV

Subscribe to Updates

Subscribe to our newsletter and never miss our latest news

Subscribe my Newsletter for New Posts & tips Let's stay updated!

About Us
About Us

Welcome to Karachi Chronicle, your go-to source for the latest and most insightful updates across a range of topics that matter most in today’s fast-paced world. We are dedicated to delivering timely, accurate, and engaging content that covers a variety of subjects including Sports, Politics, World Affairs, Entertainment, and the ever-evolving field of Artificial Intelligence.

Facebook X (Twitter) Pinterest YouTube WhatsApp
Our Picks

The world’s largest air force with the F-35 fleet in 2025

AI systems learn from many types of scientific information and run experiments to discover new materials | MIT News

Among the most troublesome relationships in healthcare AI

Most Popular

10 things you should never say to an AI chatbot

November 10, 20040 Views

Character.AI faces lawsuit over child safety concerns

December 12, 20050 Views

Analyst warns Salesforce investors about AI agent optimism

July 1, 20070 Views
© 2025 karachichronicle. Designed by karachichronicle.
  • Home
  • About us
  • Advertise
  • Contact us
  • DMCA
  • Privacy Policy
  • Terms & Conditions

Type above and press Enter to search. Press Esc to cancel.