Microsoft CEO Praises Chinese AI Model DeepSeek for Innovation

DeepSeek, an AI model developed in China, has made waves in the tech industry and significantly impacted Wall Street due to its low cost and impressive performance. Microsoft CEO Satya Nadella described DeepSeek as having "some real innovation" and emphasized that this news is "all good." These comments were made during Microsoft's recent quarterly earnings call.

Nadella asserted that AI should be viewed as any other computing development in terms of its evolution and implementation. In line with this, Microsoft announced that the DeepSeek-R1 reasoning model would be available to users of its cloud platform the same day, allowing users access to the AI model's reasoning processes.

Concerns Over Data Usage

Despite being a significant investor in DeepSeek's competitor OpenAI, Microsoft is investigating whether DeepSeek may have acquired data without authorization from OpenAI. OpenAI has claimed that they possess evidence suggesting that DeepSeek utilized their services in a way that violates OpenAI's terms of service.

However, both OpenAI and Microsoft have yet to produce any corroborating evidence for these claims. The situation has led to criticisms and allegations against DeepSeek, including statements from Howard Lutnick, a nominee for commerce secretary in the Trump administration, who accused the company of using "stolen" American technology.

Statements from Trump's Team

David Sacks, an AI adviser to Trump, commented that there is "substantial evidence" indicating that DeepSeek did indeed extract knowledge from OpenAI's models. This practice, referred to as "distillation," is described in DeepSeek's reported research but claims to be employed differently. According to DeepSeek, they trained their models like Alibaba's Qwen and Meta's Llama using the DeepSeek-R1 as the primary reference model.

DeepSeek has made its models freely available for download, allowing users with less powerful technology to operate them offline, a feature that contrasts with OpenAI's ChatGPT, where the underlying model is not accessible to users.

Identity Claims and AI Truthfulness

Public discourse has also referenced instances where DeepSeek reportedly identifies itself as ChatGPT, raising further allegations of data misuse. However, it's essential to note that AI models can misidentify themselves. For instance, a version of Google’s Gemini chatbot mistakenly identified itself as Baidu’s Ernie bot, without any claims of data theft being made by Baidu.

AI, Innovation, Microsoft