Comment on DeepSeek hype: Astounding
hype around DeepSeek. Closer examination of the background to the story shows that DeepSeek is not actually a less-than-2-year old company, but has been working on AI models since parent firm High-Flyer was formed in
2016, it
copied (distillation) OpenAI's model> so completely that, for the first week of operation, when asked "who are you?", DeepSeek answered "i'm ChatGPT.", and DeepSeek intentionally fudged their actual costs to make it look like they had created their AI for a small fraction of US AI firms' spend on development in what has to be considered a
psychological warfare tactic straight out of the CCP playbook. DeepSeek engineers did a fantastic job employing all the tweaks that US AI firms had tried before in bits & bobs, but that is hardly true "innovation", rather it's clever
adaptation of existing work done by others. Lastly, while DeepSeek's AI model is open source, which is laudable, they
haven't released any of their training process, probably because to do so would reveal the extent of IP "borrowing" they are engaged in.