The launch of the DeepSeek AI model has had a significant impact, wiping out $1 trillion in market value. Despite the initial hype, the model’s long-term effects are now under scrutiny. DeepSeek’s energy efficiency during training is notable, but its inference phase, which involves answering queries, is more complex. It uses a chain-of-thought technique, breaking down questions into logical steps, which enhances its performance in math, logic, and coding. However, this method consumes more electricity than traditional AI models. While AI currently accounts for a small portion of global emissions, there’s growing political support to increase energy allocation to AI. The debate centers on whether the energy intensity of such models is justified, particularly when used for less critical applications like generating AI content. Experts caution against overusing DeepSeek for tasks that don’t require its advanced reasoning capabilities.
Source: www.technologyreview.com















