Speaker 1: Wall Street was in chaos. $1 trillion was lost in a single day. What miracle has China done that the whole world is in chaos? AI and the tech industry are at the top of the world. And all this happened not because of a war but because of an application which China introduced. What is this application called DeepSeek? What is the reason that big tech giants have lost billions of dollars? I am your host, Dr. Nasir Vaik, on the VLC Podcast. Assalam-o-Alaikum. There has been an uproar since yesterday. Everywhere you go, you hear the same voice, DeepSeek, DeepSeek, DeepSeek. What is DeepSeek? DeepSeek is a chatbot application introduced by China. It is a software that works on the principles of AI. Just like OpenAI had created ChatGPT for you, or Google had created Gemini, or Meta and other tech giants are creating their own AI agents, AI chatbots, AI applications, AI platforms. Similarly, China has introduced an AI platform called DeepSeek which has created a ruckus all over the world. Yesterday, Wall Street fell by 1 trillion dollars. You will be surprised to know that the Wall Street that was going bullish, where tech giants were ruling, like NVIDIA, Google, Microsoft, Meta, all of them crashed. We will talk about how many of them suffered losses in the future. A Chinese named Lian Wenfeng has created this platform. Initially, it was started as a side hustle of a company called HiFlyer. Its purpose was to show people how to deliver cheap and high-quality AI. Actually, there are two things that have been considered necessary for AI technology. One is billions and billions of dollars, and the other is cutting-edge technology. DeepSeek has created such a huge ruckus that DeepSeek has broken both these rules of AI. Neither billions of dollars were needed for it, nor cutting-edge technology was used for it. How? Let me tell you. Actually, compared to the previous AI platforms which were trained by spending billions of dollars, DeepSeek has completed this platform in just 5.5 million dollars. Before that, whether it is Sam Altman's OpenAI chat from GPT, or Gemini from Google, or Meta, which is spending 60 million dollars or more, and has high hopes of spending more than that. And just a few days ago, President Donald Trump announced the Stargate project, which Sam Altman of OpenAI was present there, and he wanted to create data centers all over America, and he was saying that after this, we will take AI to a new level. Although, OpenAI has also introduced an agent, which they are calling Operator. Now, what is an agent? How does the Operator work? Even if you don't work on a computer yourself, you tell it to go and search on the internet. And you are sitting here and your mouse is moving, and it will go and search. For example, you say that I want to go to Islamabad on this weekend, and you go and see in this date where the hotel room is available in this price range, which is near a secretariat or a blue area, you give it some information like this, and it will go and start searching for it, and it will bring that information for you. If you allow it to do so much, then if you book it for me, it will book it for you as well. This type of work is now being done by an agent named Operator of OpenAI. They had introduced it, but they did not expect that they were assuming that there would be a lot of hype and the world would talk about it. But the world did not talk about it. The reason for this is that on the Apple Store, DeepSeek was the top-rated application in three countries. And those three countries were the United States of America, UK, and China. Although Apple does not have its own AI platform, but still the Apple Store decides the trend of the market. And because of this, the entire market, the entire Wall Street crashed. Because when they found out that the Chinese have made such an application, which has given such an AI platform, which is also cheap, and does all the work that Claude or Chad GPT is doing or Gemini is doing. And it is made at a much lower cost. And the biggest thing about it is that it is open source. I will tell you what all these things are in the future. First, let me tell you how many stocks of which tech giants have fallen. NVIDIA. NVIDIA's stock has fallen by 12.5%, which is a loss of $465 billion in one day. Apart from this, Broadcom has fallen by 14%. AMD's stock has fallen by 4.7%. Microsoft's stock has fallen by 4.4%. Google's stock has fallen by 3.2%. And overall, Nasdaq has fallen by 3.6%. And S&P 500 Index has fallen by 2.3%. Wall Street lost $1 trillion in one day. This is the effect of DeepSeek. It is surprising because as I told you, it has been trained for only $5.5 million. There was a reason to train for $5.5 million. Actually, something happened. America had imposed a ban on China from supplying advanced and state-of-the-art GPUs of NVIDIA. These are called graphic processing units or GPUs, which can do very complex types of processing, tasks, run algorithms, very complex types. That's why these GPUs are used in graphic designing, in videos, but GPUs are also used for AI. And NVIDIA is the company that makes these. And NVIDIA's state-of-the-art and very expensive, the more advanced, the more sophisticated, the better it is, the more expensive it will be. So those chips, all the state-of-the-art chips, America imposed a ban on them that they cannot be given to China. Now China had no choice but to use less tech apps, less tech GPUs. If they use less tech GPUs, it is obvious that they will be cheaper and their price will also be low. Now China and the people who made DeepSeek had a big challenge. That with these chips, which are not very advanced, can't run such complicated algorithms, how do we get the same work from them as JAT GPT does? Here comes the skill of their software programmers. Their software engineers actually altered the architecture of LLM. And they used small models, which were expert models. That is, they used many expert smaller models. And a central brain. Whenever you give that central brain a task, it looks at which of those small expert models is best suited to answer your question. And they used many such techniques, which made it easier for them. And because of this architectural change, it became possible for them to do the work that JAT GPT does, but with cheaper and less advanced GPUs. Actually, America wanted to maintain its control over AI. And also wanted to maintain its control. That is why it was not giving more advanced Nvidia chips to China. There is a reason for that. Come, let me show you some very interesting statistics. From 2003 to 2007, out of the 64 leading technologies in the world, America was ahead by 60. While China was only ahead by 3. Now, in today's date, in 2025, America is only ahead by 7 out of 64 leading technologies. While China is ahead by 57. That is why America fears that it is very possible that China will leave America behind in the era of technology. And that is why it wanted to maintain its control over AI. Because it knows that the world to come will be the world of AI. And apart from this, the most beautiful thing about AI is that it is an open source model. An open source model is one whose code is available to you. Its code is available on the web. If you want to make any changes in it, then you can also make those changes in it. It is up to you. And you can use it in your own way. Because it is open source, people have more confidence in it. And they think that we can use it in our own way, which is called use cases in the language of technology. It can be used for different use cases. Now, I will quickly tell you how you can install it. Actually, this model of DeepSeek is being called DeepSeek R1. And on the 20th of this month, they have launched the R1 model. Which has created a sensation in the world. How can you install it? If you want to install it, then you go to your Play Store, go to the Apple Store, and download the DeepSeek application there. But if you want to use it as an open source model, then you first download Ullama, which is a website and application. And in Ullama, then you choose the version of DeepSeek that you want to download. And after that, there is an application called Chatbox. You can download it and use its UI interface. This is a quick recipe that I am telling you. If you want to use its open source, then you go to Ullama, download Ullama. After that, you can download any version of DeepSeek that you want to use. And after that, you can use Chatbox, which is a UI interface. And after that, you will have access to its open source model. You can do whatever you want with it. If you want to make more applications, if you want to write codes, you can use it in any way you want to use it. Yesterday, when I got to know about it, I had a lot of conversations with DeepSeek. And I asked him to tell me why he was so great. So, I will tell you what DeepSeek told me about itself. And there are some things that I asked from ChatGPT, that how do you know about DeepSeek and is it really such a good application or not? It was a very interesting conversation. So, let me tell you a few things about what DeepSeek says about itself and what it tells us and what is the reason behind it. ChatGPT says that DeepSeek was launched for the first time on 15th of May, 2023, which was its version 1 and its initial release. And its purpose was to provide solutions to businesses and developers. And it was quickly accepted, because of which natural language processing and machine learning processing innovative methods were used in it. Six months later, on 20th of September, 2023, DeepSeek launched version 2, which included more than 50 languages and new features like real-time learning. And this update made DeepSeek a serious competitor for platforms like ChatGPT and Gemini. That is, it started challenging them. And after that, they released version 3, which was launched on 10th of January, 2024, in which it was further improved and modularity was added to it. What is modularity? Modularity means that you make different modules and get different things done in those modules and you can add as many modules as you want. One reason for this is that you write such a long software program, if you want to change a particular thing, then you just have to take out that module, change it and fit it in again. If you fit it in again, then the whole process continues as it is, instead of changing the whole system. That is why modular technology is considered to be very good. And you can improve it, it can also have technological advancement, and at the same time, you can also scale it, that is, you can also make it bigger. And for that, the cost of hardware is also very low. There have been more than 10 million downloads of DeepSeek. DeepSeek is telling me that I have not researched it myself. And DeepSeek says that it makes it special that its capabilities are excellent, whether it is Natural Language Processing, NLP, you are making a chatbot from it, or you are analyzing data, or you are automating customer support. DeepSeek is very friendly and provides answers in a straightforward manner. And DeepSeek also says that its modular platform-based AI design makes it a game changer. Unlike its predecessors, which came before it, it offers the same size solutions. DeepSeek gives the users the option to choose the features according to their needs. This not only reduces the cost, but also ensures that the platform is ready for specific use cases. Another outstanding feature is its multi-language support. DeepSeek can understand and create text in more than 50 languages. And what you are hearing right now, what I am telling you, DeepSeek has also told me this in Urdu. And for this reason, it is considered a great choice for global businesses. And apart from this, its real-time learning capabilities cannot be ignored. DeepSeek can immediately accept new data and user data. This ensures that it is always up-to-date and up-to-date. Actually, due to its better algorithms and open-source nature, the operational cost of DeepSeek is significantly lower. On average, businesses spend $10,000 to $50,000 annually on DeepSeek, which is affordable for them. On the other hand, OpenAI's ChatGPT spends $100,000 to $500,000 annually, which is mainly due to large datasets and expensive hardware. Gemini Google's AI platform is also expensive, and its cost can range from $200,000 to $1 million. Because it is designed for enterprise-level applications. Similarly, Meta spends $75,000 to $300,000 annually. And this is why this cost benefit makes DeepSeek popular for startups and small businesses. It allows them to use modern technology without a large AI budget. Another important factor in DeepSeek's success is its open-source nature. Unlike proprietary platforms, DeepSeek's code is freely available to the public. This sells the invention. Because developers from all over the world can participate in the platform, include new features, and make the current versions better. After this, I asked DeepSeek if it also uses NVIDIA GPUs. He replied that DeepSeek uses NVIDIA GPUs for complex computations, which provides fast and efficient operation. However, its better algorithms make it certain that it can run on low-cost hardware, which makes it accessible to a large audience. After this, I asked him that I have heard that you have created a stir in the stock market. He told me that since its launch, DeepSeek has shaken the AI industry. And there is a stir on the Nasdaq. He also told me that there is a 15% drop in the value of OpenAI, which is why investors are concerned about the competition with a cheap and open-source alternative. Because there is a lot of speculation that NVIDIA chips are so expensive and have such high demand. It is a game of demand and supply, which is why the price of a thing goes up. Until now, people thought that NVIDIA chips are very expensive and have high demand. This is why NVIDIA crashed. And OpenAI is also facing a big challenge. Because it has a competitive advantage which is available at a very low price. So, to sum up, DeepSeek is a game-changer because it is simple, low-cost, flexible, and flexible. And it connects AI platforms in a way that is not available on any other platform. And its open-source nature makes it very friendly and sells it. Whereas, its low cost makes it suitable for any business of any size. Whether you are a startup who wants to experiment, a big company that is looking for a suitable solution, DeepSeek has a lot to offer you. And with its rapid growth and adoption on a global scale, it is clear that DeepSeek is here to stay. This was the story of DeepSeek, what it has done and how it works. I have tried to give you as much information as I can in less time. West is doing a lot of research on it. I have read articles and videos of many tech experts. What do they say about it? The first thing they say is that it is very possible that the $5.5 million cost that they have set aside to train DeepSeek, it is possible that this may not be correct. Because if they say that the cost is more than this, then maybe sanctions can be imposed on them because they could not import expensive chips and GPUs. The second thing is that is DeepSeek really a revolution? Will AI not be so expensive anymore? Will it become cheaper? There is a contradiction in this. Some people say that they have programmed it in such a way that it has become modular, that AI will become cheaper. And some people say that this will not happen. You will have to use expensive chips. As I told you, OpenAI has launched an agent-operator. So when you slowly go to the higher end, you will need expensive chips. So according to them, NVIDIA is also not in danger. Donald Trump, taking the name of DeepSeek, said that it is a very good thing that technology is becoming cheaper. And now we will not have to spend billions of dollars. Maybe Sam Altman's OpenAI project, Stargate, I think it will become cheaper. And even if it does not become cheaper, then we will not have to spend 500 billion dollars on it. Let's see. Many people are saying that like every Chinese thing, it will run for a very short time and then crash or there are many problems in it. Anyway, only time will tell. Another interesting thing that the West is raising is that if you ask DeepSeek any question about China, Beijing, or about their president or about the laws that are a bit strict there, it says that it is out of my scope. They also say that this is a kind of permission that is being given by China. Now another question arises that every country in the world means to make its own chatbot which follows the rules of its censorship, which sees things from its perspective, which filters from its filters and gives information to its people. This means that just as every country has its own currency, every country has its own army, similarly, every country should have its own AI chatbot. India is working on its own chatbot. I will advise the Pakistanis that you should also start working on your chatbot. Let's make a Pakistani chatbot. I hope people who are sitting at the places where it matters are listening to me. Please do something about it. I will also tell the software engineers to work on it. AI is the future. Work on it. Build upon it. It will benefit your country and you too. It is possible that you too, whether you are 20, 25, 30 years old, young boy or girl, can become the next billionaire in the world. I would like to see that day. I hope you liked today's program. Take care of yourself. Thank you.