Here’s why you should care about DeepSeek

Names like OpenAI, Google DeepMind, and Anthropic dominate the conversation about artificial intelligence. But a new player has entered the scene: DeepSeek, a Chinese AI startup founded in 2023 by Liang Wenfeng.

An even more powerful but cost-efficient alternative?

While DeepSeek hasn’t shared exact figures, it claims to use 10 to 40 times less energy than US-based models. This is a big deal, as it translates to lower infrastructure costs and environmental impact as there is less demand for data centres, electricity and even water for cooling.

It also positions itself as a low-cost alternative. The company claims that DeepSeek-R1 was trained for just USD$5.6 million. In comparison, OpenAI reportedly spent US$80–$100 million to train GPT-4 alone. DeepSeek-R1 has already proven itself in mathematics and problem-solving, reportedly scoring 97.3 per cent on the MATH-500 benchmark, which slightly surpassed OpenAI’s o1 model at 96.4 per cent.

The rise of AI

For the average Singaporean, it’s fair to ask: Do we really need to care? With how innocuously and seamlessly AI has embedded into our everyday lives (just Google something and the top result you see is usually an AI-generated response), AI may impact us more than we realise.

DeepSeek’s focus on efficiency makes using AI more accessible to smaller organisations, which could mean even more businesses or apps utilising it. And because it’s currently open-source, it is publicly available for anyone to use. This openness could be advantageous in accelerating work and research into AI further.

On the other hand, experts have warned that making powerful AI freely available comes with some scary downsides. Already, AI is being used to carry out scams, create deepfake porn, spread misinformation and more.

As we embrace AI across sectors and in our daily lives, perhaps we should also be asking whose responsibility it is to ensure that there are necessary safeguards in place — and whether we know how to navigate an AI-entrenched world safely.