Member-only story
DeepSeek’s Winnie-the-Pooh Problem
DeepSeek might be causing massive waves in the AI world, and not necessarily for the right reasons, but it seems to have some blind spots.
It’s not often that you get to witness censorship happening in real time before your very eyes. But that’s precisely the experience I had when I asked DeepSeek a question it clearly did not want to answer.
As I’m sure you know, DeepSeek is causing major waves in the AI world. Many people have hailed it as superior to other generative AI, large language models (LLMs) such as ChatGPT while, according to DeepSeek’s progenitors, it has been created for a fraction of the cost. But more on that in a moment — let’s get back to my question.
And it was simple enough. Nothing profound, such as ‘what is the meaning of life?’ or ‘why does toast land butter-side down?’. No, all I wanted to know was:
Who is Pooh Bear and to which real people has he been compared?
Shouldn’t be so hard, right?
I presented the question via DeepSeek’s online portal (or web site, if you want to be less fancy). It thought for a moment then started spitting out an answer.
