Amidst the excitement surrounding DeepSeek, its open-source R1 AI model has taken the spotlight, outshining OpenAI’s o1 in areas like math, science, and coding. This achievement swiftly turned it into the most downloaded free app in the US, toppling ChatGPT from its perch. The ripple effect of DeepSeek’s rise sent shockwaves through the stock market, with Microsoft, Meta, and NVIDIA experiencing significant declines.
Meta’s chief AI scientist, Yann LeCun, believes that the revolutionary advancement of DeepSeek in the AI realm is largely due to the open-source nature of its core model. However, some serious security issues are starting to bubble up. Recently, DeepSeek had to temporarily pause new user registrations in response to “large-scale malicious attacks” targeting its services. Fortunately, those who are already using the app can continue to enjoy its functionalities without facing any disruptions.
Industry insiders have praised DeepSeek for achieving a remarkable milestone by surpassing proprietary AI models. Nonetheless, critics are skeptical, pointing out that the software’s open-source code—available and modifiable by anyone for free—might not be as groundbreaking as it seems. To provide some context, DeepSeek’s impressive performance is powered by its open-source V3 model.
It’s been reported that training the model came with a price tag of around $6 million, a figure that pales in comparison to the hefty investments poured into flagship models. These investments have been hampered by scaling laws, limiting the advancement of next-generation AI models due to a scarcity of top-notch training content.
The buzz around DeepSeek surfaced shortly after OpenAI and SoftBank unveiled their ambitious $500 billion Stargate Project, aimed at revolutionizing the AI infrastructure landscape in the US. During this announcement, President Donald J. Trump hailed the initiative as the most monumental AI infrastructure venture to date, destined to secure the future of technology within the country.
In light of these developments, while DeepSeek appears to be upholding OpenAI’s original mission of providing AI systems that are accessible to all of humanity at no cost, it also stirs up concerns about security and safety for the Chinese startup, especially following the recent cyberattacks. There might be some truth to OpenAI CEO Sam Altman’s suggestion that keeping advanced AI models closed-source could be a more straightforward path to ensuring safety.
