DeepSeek AI: Revolutionizing the Cost of Generative AI
In the rapidly evolving world of artificial intelligence, a new player has emerged, capturing the attention of tech analysts and industry insiders alike. DeepSeek AI, a startup based in China, has made headlines for its innovative approach to training generative AI models. This breakthrough has the potential to significantly reduce the costs associated with developing AI platforms, such as OpenAI’s ChatGPT and Google’s Gemini Ultra.
DeepSeek AI’s rise to prominence is largely due to its ability to train AI models more efficiently than its competitors. According to a report by Ben Thompson on his website Stratechery, DeepSeek was developed under unique constraints that forced the team to innovate. The company operates under a trade embargo that limits access to high-quality semiconductor chips from Nvidia, a leading American multinational corporation. As a result, DeepSeek’s developers had to rely on lower-quality chips, which led them to incorporate a variety of AI optimization techniques to maximize efficiency.
The financial implications of DeepSeek’s approach are staggering. The developers claim that training their latest AI version cost approximately $5.6 million, a fraction of the $78 million reportedly spent on training ChatGPT 4 and the $191 million for Google’s Gemini Ultra, as noted in Stanford University’s 2024 Artificial Intelligence Index report.
This cost reduction is not only a boon for the developers but also for scientists and consumers. Umar Iqbal, an assistant professor of computer science and engineering at Washington University in St. Louis, highlighted the potential benefits. His lab, like many others, spends tens of thousands of dollars to access AI platforms. The competition introduced by DeepSeek could drive prices down, making AI technology more accessible.
One of the key methods employed by DeepSeek to reduce training costs is “distillation.” This technique involves using an established generative AI system, such as ChatGPT, to “teach” their system how to perform tasks. PhD students from McKelvey Engineering have recently explored distillation to enhance large-language models without additional training.
“For technologies to be mass adopted, they need to be cheap,” Iqbal stated. “What this has demonstrated is using models can become very cheap. Overall, this is an interesting development. It significantly reduced the cost of AI. We’ll be able to conduct experiments, more large-scale experiments.”
Concerns with DeepSeek
Despite the promising advancements, Iqbal also raised concerns about the potential pitfalls associated with DeepSeek. One major issue is the need for large-scale hardware to run these models, which is not something that can be easily downloaded onto a personal device. The AI platforms operate by connecting a user’s machine and data to the AI machine in the cloud, which can lead to a loss of control over personal data.
“And that is a very serious concern,” Iqbal emphasized.
AI systems have the capability to create vast surveillance infrastructures, some of which are already in place through search engines that track user data for e-commerce purposes. “All this data will go to different AI vendors, and they can use that information to profile users, infer their interests, surveil them and maybe also influence them,” Iqbal explained.
The location of DeepSeek’s headquarters in a country with strict governmental oversight further complicates matters, raising questions about how user data is managed. “This creates a serious big security and privacy safety risk,” Iqbal warned.
Another concern is the increasing integration of AI language models into mobile apps. AI is being used for various applications, such as planning vacations. However, if any malicious software is present, it can potentially harvest more data from the user and manipulate the AI’s results.
“The technologies, when they have a lot of potential, they evolve very quickly,” Iqbal noted. “You need to have guard rails and protections buried into the design. This is not happening with AI systems.”
- DeepSeek AI’s innovative approach could lower AI development costs.
- Concerns about data privacy and security remain significant.
- Integration of AI into mobile apps poses additional risks.
As DeepSeek AI continues to make waves in the industry, it is clear that its impact will be felt across various sectors. While the cost-saving benefits are undeniable, the challenges related to data privacy and security must be addressed to ensure the responsible and ethical use of AI technology.
Originally Written by: Leah Shaffer