Deepseek Test: First Impressions

Deepseek, a low-budget ChatGPT, is currently making waves by sparking debates on technology and the impact of AI on Wall Street. Despite its budget limitations, this AI has garnered growing interest thanks to its intriguing, though improvable, capabilities.

During testing, Deepseek was tasked with several key points.

Regarding information retrieval, it effectively responded to a request for a plan to improve traffic on a site dedicated to generative AI. Its response was structured and relevant, well-developed to provide a solid foundation of ideas.

However, concerning news retrieval, I asked it to provide the latest on ChatGPT, and it failed to identify the announcement of the StarGate plan, a significant event for the tech community. While the information provided was plentiful, it wasn't always relevant, highlighting the current limitations of its algorithm in retrieving specific news.

In terms of writing, the quality of the response was interesting. The text produced was well-developed and adhered to the instructions given in the prompt, reflecting a good content generation capability.

As for document analysis, Deepseek showed shortcomings. It cannot analyze images and required OCR-processed documents for in-depth analysis. When importing OCR-processed PDF and Excel files, bugs were encountered, making it difficult to fully assess its analysis capabilities.

In conclusion, although Deepseek offers interesting first impressions, improvements are needed, particularly in document analysis and news retrieval. These aspects will be explored further in an upcoming article for a more comprehensive evaluation.