DeepSeek has taken the tech world by storm, riding on a wave of excitement thanks to the stellar performance of its open-source R1 AI model. It not only outshines the reasoning capabilities of OpenAI’s o1, particularly in math, science, and coding but has also leapfrogged over ChatGPT to become the most downloaded free app in the United States. In light of this, the stock prices of giants like Microsoft, Meta, and NVIDIA have seen a significant dip, driven by the market frenzy DeepSeek has sparked.
On the other hand, while DeepSeek has been celebrating its achievements, some serious security alarms have begun to ring. Yann LeCun, the lead AI scientist at Meta, attributes DeepSeek’s success to the open-source nature of its model. However, the company has hit the brakes on new user registrations, citing “large-scale malicious attacks” that have targeted its services. Despite these troubles, current users can still access all the features without any glitches.
Industry leaders are impressed by how DeepSeek’s AI has managed to surpass other proprietary AI models. Yet, some critics argue that this accomplishment might be a bit overstated because of the software’s open-source code, which is available for free tweaking and use by anyone. DeepSeek’s app gets its power from their open-source V3 model, a fascinating piece of tech.
What’s even more intriguing is that this model was trained on a relatively modest budget of about $6 million. This figure is a mere drop in the ocean compared to the vast sums being funneled into flagship models, especially when considering the bottleneck these models face with a lack of high-quality training content.
This enthusiasm around DeepSeek coincides with OpenAI and SoftBank’s colossal $500 billion Stargate Project, aimed at building state-of-the-art infrastructure to drive AI progress across America. In an announcement, President Donald J. Trump touted this initiative as the largest AI infrastructure endeavor ever, aiming to “keep the future of technology” firmly anchored in the U.S.
Amidst this landscape, DeepSeek continues to walk the fine line of OpenAI’s original mission, which is to freely provide AI systems that serve humanity. Nevertheless, the recent wave of cyberattacks raises pressing questions about security, posing a potential challenge for the Chinese startup. OpenAI’s CEO, Sam Altman, might have a point about keeping advanced AI models under wraps to meet safety benchmarks more easily.
