[ad_1]
DeepSeek: Open source success
| Photo Credit:
Dado Ruvic
It has been more than two years since ChatGPT’s debut upended the Artificial Intelligence (AI) world and led to massive global investments in AI technologies and infrastructure, but the world was once again turned upside down with DeepSeek’s emergence.
The essence of the DeepSeek moment is that it demonstrated training an advanced Large Language Model (LLM) was possible at lower costs than previously imagined, which sparked a market panic and sell-off of AI infra companies’ stocks due to anticipated slowdown of revenue. But the opposite is likely to happen.
Inflection point
When many other major technologies like Wi-Fi and 5G were first introduced, they went through a pattern of initial hype, a subsequent period of scepticism, but eventually became mainstream ; the same thing is happening to AI. The DeepSeek moment is an inflection point that will accelerate AI adoption and innovations, leading to demand for AI infrastructure, but in a frugal way.
The DeepSeek moment made people realise that training LLMs is not just the domain of companies that had the resources to purchase large numbers of expensive graphics processing units (GPUs).These LLMs can be deployed on less resource-intensive Information Technology (IT) infrastructure and even existing IT infrastructure.
Technically any chip can run AI, but choosing the right tool for the right job after considering cost, performance, and efficiency is what will determine if a business needs more GPUs, Central Processing Units (CPUs), Field Programmable Gate Arrays (FPGAs), or other AI accelerators.
Flexibility
Businesses thus have the flexibility to choose the AI infrastructure that best meets their goals.Another key factor that makes DeepSeek stand out is its open-source approach, when most AI models today are closed source. While DeepSeek is not the first or only open-source LLM, its success has certainly shone the spotlight again on how an open-source model accelerates the accessibility of AI.
Open source allows anyone to download, copy, and build upon the model’s source code. Not only can anyone scrutinize the AI models, but it also enables startups and developers with fewer resources to access cutting-edge AI without licensing fees, allowing them to build and distribute their own AI applications by leveraging the same technologies.
Open source
I believe that open source makes AI development more accessible, provides choice, and leads to expansion for both AI innovations and AI infrastructure demand.
With the rapid development of AI, as seen through examples like more advanced reasoning models that require even more compute power, there is no doubt that both open and closed source systems will continue to see growth in the near future.
In the AI space, a closed ecosystem limits technology to the haves and not the have-nots. It reduces customizations for developers to pivot; slows down innovation as developers need to rely on the providers for updates, access, and support; and most importantly, does not offer transparency for others to inspect the datasets behind the models to ensure ethical and safe use of AI.
The writer is Vice President and Managing Director of Intel, India Region
Published on April 3, 2025
[ad_2]
Source link
