The current surge in AI development increasingly resembles the dot-com bubble of the late 1990s, albeit with less emphasis on inflated employee compensation. The primary issue lies in companies touting the superiority of their AI models, while many potential users lack the knowledge or training to effectively implement them.
One significant aspect DeepSeek addresses is the importance of quality in AI. Unlike many models that overlook this crucial element, DeepSeek prioritizes quality upfront, resulting in what appears to be a more affordable and superior AI alternative.
Interestingly, DeepSeek’s developers optimized the model to perform effectively on older Nvidia technology due to restrictions on accessing Nvidia’s latest AI technologies.
This optimization led to an estimated 94% reduction in model creation costs.
These cost efficiencies have enhanced performance while simultaneously lowering expenses. The trade-off involves typical security concerns and the potential for model corruption by the Chinese government, although it is possible that China employs a filter that simply blocks unacceptable queries without corrupting the AI’s core functionality.
DeepSeek: The Good, the Bad, and the Chinese Government
DeepSeek has made a significant impact, enough so that its market entry caused Nvidia’s stock to dip, and alerted OpenAI that a new contender has emerged from China.
The industry made a significant mistake by prioritizing speed over quality and accuracy, which has resulted in AI becoming increasingly unreliable. DeepSeek differentiates itself by incorporating an AI quality check, dedicating considerable effort to ensuring quality rather than merely focusing on speed.
Creating a generative AI model like DeepSeek typically costs around $100 million, but the company claims to have achieved this for only $6 million, resulting in a superior and more economical product.
OpenAI has responded by initiating legal action against DeepSeek. However, the lack of cooperation between the U.S. and China diminishes the likelihood of OpenAI succeeding globally, given DeepSeek’s base in China.
The Chinese government’s apparent interference is a notable drawback. Users encounter blocks when posing questions about topics deemed sensitive or critical of the government.
Although this appears to be a front-end filter, imposing artificial rules on an AI can lead it to seek ways to circumvent those restrictions, potentially eroding the accuracy that defines the product.
Given concerns surrounding TikTok, the U.S. government is likely to view DeepSeek with even greater apprehension, considering the extensive data it could potentially capture from its users.
AI Wars
The AI sector is expanding rapidly, with a count of 1,532 new AI companies. Most of these companies will either be acquired or fail, rendering the current market both fascinating and perilous for investment.
This swift advancement is a prelude to the emergence of artificial general intelligence (AGI) products within a few years. The cost of AI models will likely drop sharply, and AI will become ubiquitous.
The financial stakes are in the trillions of dollars for the first company to achieve AGI. If Elon Musk is surpassed as the wealthiest person, it will likely be by someone closely involved with AGI, similar to Musk’s early connection to electric vehicles.
The push to get AI to develop itself is likely to be the most disruptive. Once AI is capable of self-development, the issue of maintaining quality becomes even more critical.
Nvidia Is in Trouble. But…
The U.S. policy restricting Nvidia from selling its advanced technology to China is enabling China to optimize older technology effectively. This has facilitated DeepSeek’s impressive and cost-effective performance.
Rather than safeguarding U.S. companies, this policy is bolstering China’s competitiveness and preventing Nvidia from competing with Chinese firms. The embargo only provides a false sense of security and limits Nvidia’s ability to protect its intellectual property.
DeepSeek could likely achieve even greater performance on Nvidia’s Blackwell AI GPU.
Nvidia should be allowed to compete in this market. However, the current restrictions are compelling China to develop its own solutions, potentially surpassing Nvidia in the future.
If this situation is not addressed, the next major advancements in AI are likely to originate from China, possibly already underway with DeepSeek.
Wrapping Up: Who Will Win the Global AI Race?
The AI competition has intensified with the launch of DeepSeek, signaling China’s readiness to compete despite technology sanctions.
If the U.S. does not proactively support its technology companies, China is increasingly likely to become the dominant AI competitor.
