The contribution Hype around China-Ki: Can Deepseek Chatgpt give the water? Fabian Peters first appeared on Basic Thinking. You always stay up to date with our newsletter.
Feeled like out of nowhere, the Chinese company Deepseek launched a serious alternative to Chatgpt. The AI model should be cheaper and more energy-efficient than its model from Openaai. The US exchanges therefore faltered. But can Deepseek Chatgpt really be enough? An analysis.
The Chinese company Deepseek has put the AI industry into turmoil. Because the company of the same name of the company of the same name has conquered the Apps Stores in the storm overnight. In the United States, the Deepseek app even overtook Chatgpt and in the meantime was number one of the most downloaded apps for the iPhone.
The background: The AI model “Deepseek R1” can not only keep up with the large voice models from Google, Meta and Openaai, it trumps them official information even partially. It seems particularly impressive that Deepseek should be significantly more cost -effective and energy -efficient with the same performance.
Immediately after its presentation, the AI confused the financial markets. Numerous well-known technology companies experienced a raven-black day on the stock exchange.
What is Deepseek?
Deepseek is a Chinese start-up that specializes in the development of voice models and artificial intelligence. The company’s origins go back to a AI-based hedge fund from 2015.
In April 2023, the independent AI company Deepseek emerged from the financial offer, which has since been committed to the development and marketing of voice models based on artificial intelligence.
After several tentative attempts and publications, the “Deepseek R1” model attracted internationally. Because the AI should Company According to it, not only can keep up with established AI systems such as Chatgpt, but also even exceed them.
On top of that, Deepseek R1 is cheaper and more energy -efficient in the company. In combination with the conquest of the app charts, they provided snapshot on the stock exchanges. In the meantime, numerous renowned tech companies lost five to ten percent of stock value-especially in the United States.
Including: Nvidia (minus 10 percent), Microsoft (minus 6 percent), alphabet (minus 4 percent) as well as Meta and AMD (each minus 5 percent each. The Dutch chip manufacturer ASML also recorded a full minus in Germany with around ten percent. Above all, Infineon and Siemens with a minus of four percent each.
How does Deepseek R1 work?
Deepseek R1 is designed for complex tasks and solutions. The voice model was published under the so-called co-license of the Massachuchet Institute of Technology. Based on an unreserved open access, this promotes and allows both academic and commercial use – without restrictions.
Deepseek consciously relies on a far less restrictive approach than many other AI providers. The aim of the company is to stand out from the competition by transparency and save costs. In principle, Deepseek R1 is available free of charge both private users and companies. A basic version allows general use via app or desktop.
German is officially supported, but Deepseek is not yet completely powerful for the language. Therefore, errors can occur in the German version. The reason: The AI model was primarily developed for the Chinese and US market.
For professional applications, Deepseek offers a paid license model. It includes additional functions. Among them: your own data storage and additional language packages. The language model is aimed primarily at companies. Deepseek R1 is a AI-based dialog system that builds on large voice models.
Compared to other models, the AI should need significantly less computing power for the same performance. Deepseek therefore has a lower need for powerful AI chips, which saves costs and makes the system more energy-efficient.
Like other chatbots, the AI is based on so -called neuronal networks, the language broke into context and meaning. In addition to public content, Deepseek also includes other sources such as specialist publications or official documents. The model should learn during operation. Through continuous training, it always remains up to date.
criticism
With regard to its sources, Deepseek relies on transparency. The voice model refers to any uncertainties in information in the dialogue or even refuses answers if sources are not clear. This should make the chat bot more precise than other AI models.
The Official API However, Deepseek R1 targeted certain content. This applies above all to topics that are politically delicate in China. For example, the AI regularly refuses to answer questions about the suppression of Uighers or human rights violations within the People’s Republic.
Due to the open and customizable open source approach from Deepseek, this census tower mechanism can theoretically be levered out, but only with great effort. Deepseek founder Liang Wenfeng also maintains close connections to the Chinese government and the Communist Party.
Critics fear that the AI model could be used to operate espionage, to spread false information or to influence the policy of other countries. Similar accusations are also raised compared to the Chinese short video platform Tiktok.
Deepseek and Chatgpt in comparison
Deepseek continuously feeds his voice model with current data records. In contrast, Chatgpt mainly uses specified information from a certain period of time – at least in the basic version. According to experts, the Deepseek R1 model can score with its performance. Some call the open source approach as a gift to the world.
Different performance tests have shown that the AI system transfers the Chatgpt O1 model, which is considered the “most intelligent” on the market, into several disciplines. According to Deepseek, just five percent of Openaai costs. This and the fact that the Chinese company was able to develop a competitive chat bot in just a few months is currently worried about the AI industry.
While Chatgpt tends to cover everyday questions, Deepseek is primarily the focus. Including: medical, legal and technically complex topics in which correct and reliable information is essential. In contrast to Chatgpt, Deepseek therefore points out potential mistakes.
According to experts, Deepseek R1 can only partially keep up with other large AI models. The company puts the areas in which it is better in terms of performance. Especially due to the cost advantages, the AI is a serious competitive offer. However, there is also a misunderstanding behind it.
Costs and misunderstandings
Some investors were particularly amazed at the fact that the training of the Deepseek AI is said to have cost just around $ 5.5 million. For comparison: Openai invested almost $ 100 million in the training of Chatgpt-4. However, this comparison lags.
Because Deepseek himself restricted that the sum only applies to the official training of Deepseek-V3. So there are no additional costs as prior research, experiments or the development of algorithms. Another reason for the relatively affordable prices is the company’s open source approach.
For example, customer can operate their own servers, which provides you with data sovereignty and significantly reduces the costs for deepseek. The lower prices also come about because the company always extracts smaller AI models from larger ones. In addition, Deepseek has no functioning business model so far. Specifically, the company is currently not making a profit, but even loses money.
Deepseek nevertheless proves that AI systems can be operated much more cheaply and more energy-efficient, which should lead to a rethink in the competition. The same applies to the company’s open source approach. In comparison, Openai pursues an extremely restrictive approach. However, Deepseek has not used or developed any revolutionary technologies for its voice model, but only optimized the approaches of its competitors.
The company is likely to mix the AI industry and big-tech and get back to the bottom of the facts-this affects the USA in particular. The current hysteria on the stock exchange still appears overdoing.
Also interesting:
- AI project “Stargate”: USA put Europe under pressure-despite unclear financing
- “Free Our Feeds”: New initiative should protect social media from billionaires
- Fraud on Twitter takeover? US stock exchange supervision sues Elon Musk
- Electricity and water consumption: the effects of AI on the environment
The contribution Hype around China-Ki: Can Deepseek Chatgpt give the water? Fabian Peters first appeared on Basic Thinking. Follow us too Google News and Flipboard.
As a tech industry expert, I believe that the hype around China-Ki and Deepseek Chatgpt is warranted to some extent. Both China and Ki are known for their advancements in technology and artificial intelligence, so the collaboration between the two countries could lead to some exciting developments in the field.
However, it is important to approach this hype with caution and not get too caught up in the excitement. While Deepseek Chatgpt may have the potential to revolutionize the way we interact with technology, it is still in its early stages and there are likely to be limitations and challenges that need to be overcome.
In terms of giving the water enough, it will be important for the developers of Deepseek Chatgpt to ensure that the technology is ethically and responsibly implemented. This includes addressing concerns around data privacy, bias in AI algorithms, and the potential impact on society as a whole.
Overall, I am optimistic about the potential for China-Ki and Deepseek Chatgpt to make a significant impact in the tech industry, but I also recognize the need for careful consideration and oversight as these developments unfold.
Credits