Yazılar

Hugging Face Works on Fully Open-Source Alternative to DeepSeek-R1 AI

/in /tarafından

Hugging Face has launched a new initiative to develop Open-R1, a fully open-source replication of the DeepSeek-R1 AI model. This move comes in response to last week’s release of DeepSeek-R1 by the Chinese AI firm DeepSeek, which made headlines for its advanced capabilities and potential to rival OpenAI’s cutting-edge models. While DeepSeek-R1 was made publicly available, it was not truly open-source, as crucial components like the training code and dataset were withheld. Hugging Face aims to bridge this gap by reconstructing these missing elements, ensuring a fully transparent and accessible alternative for the AI community.

Why Is Hugging Face Building Open-R1?

In a blog post, Hugging Face researchers outlined their motivation for replicating DeepSeek-R1. While the model’s architecture and weights were shared, key training assets were not disclosed, making it a “black-box” release. This means users can run the model locally, but they lack the necessary data and methods to recreate or modify it. By developing Open-R1, Hugging Face hopes to empower researchers and developers with a fully open framework, promoting transparency and collaborative AI advancements.

One of the critical missing pieces in DeepSeek-R1’s release is the dataset used for training, particularly in reasoning-specific tasks. Additionally, the training code that defines hyperparameters—essential for fine-tuning the model’s ability to process complex queries—remains undisclosed. Hugging Face’s initiative aims to reconstruct these elements, ensuring that developers can understand and improve upon the model rather than simply using it as a locked-down tool.

By working on Open-R1, Hugging Face is reinforcing its commitment to truly open AI development, countering the growing trend of AI models being released with limited transparency. If successful, this project could set a new standard for open-source AI, allowing researchers to study, improve, and build upon state-of-the-art models without restrictions. As AI development continues to accelerate, efforts like Open-R1 will be crucial in maintaining a balance between innovation and accessibility.

Microsoft Investigates Possible Unauthorized Access to OpenAI Data by DeepSeek-Linked Group

/in /tarafından

Microsoft and OpenAI are conducting an investigation into whether a group associated with Chinese AI startup DeepSeek improperly accessed OpenAI’s data. According to sources familiar with the matter, concerns arose when Microsoft’s security team detected unusual activity involving OpenAI’s application programming interface (API). The group in question allegedly extracted large amounts of data in a manner that may not have been authorized, prompting further scrutiny from both companies.

The suspicious activity was first observed in the fall when Microsoft researchers noticed individuals believed to be linked to DeepSeek transferring significant volumes of data via OpenAI’s API. While OpenAI allows developers to license its API to integrate its AI models into their own applications, excessive data extraction could indicate an attempt to bypass OpenAI’s built-in usage restrictions. If confirmed, such actions may violate OpenAI’s terms of service, raising legal and ethical concerns over the security of proprietary AI models.

DeepSeek recently introduced its own AI model, R1, an open-source system that claims to rival or surpass leading AI models from OpenAI, Google, and Meta on key industry benchmarks. The model, designed to replicate human reasoning, has positioned DeepSeek as a formidable competitor in the AI sector. Notably, R1 was developed at a fraction of the cost of its Western counterparts, further intensifying competition in the rapidly evolving field of artificial intelligence.

The potential unauthorized access and the emergence of a strong competitor have already had significant market repercussions. Following the news, AI-related stocks, including Microsoft, Nvidia, Oracle, and Google’s parent company, Alphabet, saw a sharp decline, collectively losing nearly $1 trillion in market value. As Microsoft and OpenAI continue their investigation, the case underscores growing tensions in the AI race, particularly as global competition heats up between U.S. tech giants and emerging players from China.

Sberbank Plans AI Collaboration with China Amid DeepSeek’s Rise

/in /tarafından

Sberbank, Russia’s largest bank, is set to collaborate with Chinese researchers on artificial intelligence (AI) projects, a top executive revealed, following the groundbreaking success of China’s DeepSeek AI model. DeepSeek, which has created a highly cost-effective AI model, has shaken up the tech landscape, challenging the dominance of U.S. companies like Nvidia. This move reflects the growing collaboration between Russia and China, which share a “no limits” strategic partnership, particularly in the realm of AI.

Under CEO German Gref, Sberbank has transitioned from its bureaucratic past as a Soviet-era state savings bank into a leader in AI innovation. In 2023, the bank launched its own AI model, GigaChat, and is now focusing on joint research projects with China. Alexander Vedyakhin, Sberbank’s First Deputy CEO, confirmed the plans but refrained from naming specific Chinese research partners. The partnership is expected to strengthen the scientific ties between the two countries, in line with Russia’s broader push to expand cooperation with China in areas such as military AI applications.

DeepSeek’s innovative, low-cost AI models have caused a stir globally, with investors now viewing them as serious competition for U.S. tech giants. The potential Russia-China AI alliance could further disrupt the global AI sector, especially as the race between China and the U.S. for AI supremacy intensifies. The shared vision of Presidents Vladimir Putin and Xi Jinping, who view the West as in decline, could fuel further collaboration in emerging technologies like AI, quantum computing, and synthetic biology.

Despite facing sanctions, Russia is eager to build AI partnerships with China and other BRICS nations to counterbalance U.S. dominance. However, Russia’s own AI progress remains difficult to gauge, as some projects are classified. Additionally, both Russia and China struggle with limited domestic computing power due to sanctions, prompting efforts to seek alternative methods to stay competitive in the AI race.

Sberbank’s comparison of its GigaChat MAX model to DeepSeek’s offerings reveals that while DeepSeek excels in scientific tasks, GigaChat remains competitive in the banking sector. Vedyakhin acknowledged DeepSeek’s success as proof that quality AI models can be built without massive investments in infrastructure, such as the U.S. megaproject Stargate. Sberbank, like DeepSeek, has made most of its AI platforms publicly accessible, including its text-to-image model, Kandinsky, and GigaChat Lite, which reflects a more transparent approach than that of OpenAI, drawing a large community.