DeepSeek—Interview With A Chatbot
You may have heard about the Chinese chatbot, DeepSeek. It made news this past month which spooked investors and tanked the tech stocks on Jan 27. Led by Nvidia, the NASDAQ fell more than 3%.
DeepSeek's large language model (LLM) is basically a cheaper, more efficient ChatGPT. The developers used far fewer chips than any other leading chatbot and buit DeepSeek at a fraction of OpenAI’s budget for ChatGPT. They claim to have used just 2000 Nvidia chips which they acquired prior to when US export sanction went into place. By comparison, OpenAI used 25,000 of the more powerful Nvidia H100 chips to build GPT-4.
Translated: The cost to built DeepSeek was less than $6 million verses $100 million for GPT-4. And that's why investors in US tech are spooked. They are suddenly concerned about the competitive edge of U.S. AI companies.
So, is DeepSeek any good?
Looking similar and functioning like ChatGPT, DeepSeek offers a free alternative to OpenAI's chatbot. And, since I'm a ChatGPT power user, I figured I should find out what DeepSeek can actually do. So I interviewed it.
Question: can you code an HTML document for me if I give you the raw content?
DeepSeek: Absolutely! If you provide the raw content, I can help you code it into a properly structured HTML document. Just share the text, images, or other elements you'd like to include, and let me know if you have any specific design preferences (e.g., headings, lists, tables, styles, etc.). I...
