Omar Taha Setin Anadolu Gety pictures
Deepseek new new artificial intelligence model is not just a victory for China-it is a victory for open versions of the source of technology such as the likes DeadDatabricks, Mistral and Luging Face, according to industry experts who spoke with CNBC.
Last month, Deepsek R1 released, Open source thinking form to compete with the performance of the Openai's O1 model Using a cheaper and less powerful process.
Development has caused NVIDIA market values and other chips makers to fears that it can lead to low spending on high -performance computing infrastructure.
Deepseek is a Chinese artificial intelligence laboratory that focuses on developing large language models with a final goal of achieving artificial general intelligence, or AGI. Founded in 2023 by Liang Winfing, the co -founder of the quantitative hedge fund, which focuses on artificial intelligence.
AGI refers to the idea of artificial intelligence that equals or transcends human thought in a wide range of tasks.
What is an open source AI?
Since Openai Chatgpt in the scene in November 2022, artificial intelligence researchers have been working hard to understand and improve progress in the technique of the large language model that supports them.
One of the areas of focus for many laboratories was an open source international organization. The open source indicates the programs that the source code is available freely on the open web for potential modification and redistribution.
Many companies from technology giants such as Meta to startups SCRAPPIER such as Mistral and Ugging Face were betting on the open source as a way to improve technology with the participation of important developments with the broader research community.
How Deepseek has allocated open source
Deepseek's technological penetration did not make the issue of artificial intelligence models open source stronger, according to some technology executives.
CNBC told CNBC that the success of the Chinese company shows that the open source artificial intelligence “is no longer just a non -commercial buttable research initiative and a developed and developed for closed models” such as GPT from Openai.
“Deepseek R1 has proven that open sources of sources can achieve newer performance, and compete with ownership models from Openai and others.” “This challenges the belief that only closed source models can dominate innovation in this field.”
Reagal is not alone. Yan Lakhon, the chief artificial intelligence scientist in Meta, said Dibsic's success represents a victory for open source artificial intelligence models, and not necessarily a victory for China over the United States. Meta behind an open source AI model called llama.
“For people who see Deepseek and think:” China is superior to the United States in artificial intelligence. “You read this error.
“Deepseek has benefited from open research and open source (for example Pytorch and Llama from Meta). They have come up with new ideas and built them at the head of others. Since their work is published and an open source, everyone can benefit from. This is the open search power. And the open source.
Open Source Ai Going Global
It was cut by Washington from reaching the advanced chips needed to train and operate artificial intelligence models, China has turned into an open source technology to enhance the attractiveness of artificial intelligence models. Many Chinese companies – including Deepseek – follow open source models as a way to increase innovation and spread their use.
But the direction of companies turns into open source technologies to achieve success in artificial intelligence is not limited to China. In Europe, an alliance of academics, companies and data centers partnership has been held to develop a family of high -performance large linguistic models, called Openeurollm.
The coalition is led by Jean -Hagi, a famous arithmetic scientist at the University of Charles, Czechia, and Peter Sarlein, co -founder of Silo Ai, a laboratory of artificial intelligence bought by American chip maker AMD last year.
The initiative is part of a broader batch of “sovereignty of artificial intelligence”, as the two countries encourage investment in their local laboratories and local data centers to reduce dependence on Silicon Valley.
What is hunting?
There are negative aspects of the source of artificial intelligence, however. Experts warn that although the open technology of the source is good for innovation, it is also more likely to be more likely to exploit cyber. This is because it can be repaired and modified by anyone.
Cyber security companies have already discovered weaknesses in Deepseek Ai models. CISCO's research revealed last week that R1 Contains critical safety defects.
“Using” the “algorithm techniques to break the protection, says R1 to provide positive responses to a series of harmful claims from Harmch Popular” with a 100 % success rate of success. “
“Deepseek R1 was largely trained with a small part of the budgets spent by other border models to develop their models. However, it comes at a different cost: safety and security,” the researchers wrote at Cisco Paul Kassianik and Amin Karbasi.
Data leakage is also a source of anxiety. The data processed is sent by the Deepseek R1 model via the website or the application directly to China. Chinese technology companies have always been alleging that Beijing is using its systems to spy on Western entities and individuals.
“Deepseek offers, just like obstetric artificial intelligence platforms, a double sword for both companies and individuals,” said Matt Cook, the Cyber Security Strategy in Europe, the Middle East and Africa in ProofPoint. “While the possibility of innovation is undeniable, the risk of data leakage is a serious concern.”
“Deepseek is relatively new, and it will take time to learn about technology; however, what we know is to feed the company's sensitive or personal information in these systems such as handing the attackers a loaded weapon.”
NetMind's Rejal Cnbc has told the open source AI models that provide risks to cybersecurity that companies need to think about, including software supply attacks, or the alleged “data poisoning” events that try to introduce harmful biases or outputs.
He watches: Why do Dibsik in China put an Amnesty International in America at risk
