DeepSeek, a rising star in the AI industry hailing from China, has certainly captured the spotlight, largely thanks to its impressive R1 model. This model is making waves by outperforming OpenAI’s well-known o1 reasoning model in areas like math, science, and coding—all while costing just 3% of what competitors spend.
What sets DeepSeek apart is its commitment to making its advancements accessible. Unlike the behemoths like Microsoft and OpenAI, which depend on significant resources and intensive training, DeepSeek’s model is open-source. This means anyone interested can delve into and duplicate its development journey without incurring the hefty costs often associated with high-end AI models. Tests have shown DeepSeek’s model outpaces leading contenders such as GPT-4o, Gemini, and Claude.
Naturally, this has stirred quite a buzz within the AI community. Experts and enthusiasts alike are stunned by the model’s remarkable performance combined with its economical approach. Even Yann LeCun, the leading AI scientist from Meta, has weighed in on the significance of DeepSeek’s achievement. Speaking to Business Insider, he expressed:
“They took existing ideas and innovated upon them in meaningful ways. Their decision to publish and make their work open source means everyone benefits. That’s the true power of open research and sharing knowledge.”
The ripple effect of DeepSeek’s innovation has cast doubt on the US’s dominance in the AI arena. This comes amid just-announced plans by OpenAI and SoftBank to invest $500 billion in building high-level AI infrastructure across the US, dubbed the Stargate project. However, following this news, stocks saw a noticeable dip, highlighting the challenges posed by DeepSeek’s progress.
LeCun believes DeepSeek’s breakthrough transcends typical US-China competitiveness, shedding light on the value of open-source models. It’s a poignant reminder of OpenAI’s original mission, which embraced openness. He even suggests that “open-source models might eclipse proprietary ones.”
DeepSeek’s R1 stands tall among the few open-source heavyweights, like Meta’s Llama. Interestingly, last year, OpenAI CEO Sam Altman mentioned plans for future open-source offerings, despite the company’s shift from its original open-source ideals. He has since argued that keeping advanced models proprietary helps maintain safety standards.
Recently, DeepSeek launched an app that took the market by storm, quickly outpacing the widely-loved ChatGPT as the top free app in the US. Meanwhile, as DeepSeek garners attention, giants like Nvidia, Microsoft, and Meta have felt the impact, with their stock prices taking a hit following the company’s surge in popularity.
