Deepseek thought for 19 seconds before answering the question, “Are you smarter than Gemini?” Then it delivered a whopper: Deekseek thought it was chatgpt.
This seemingly innocent mistake could be proof – a smoking gun per Says – yes, Deepseek was trained on Openai models, as it has been claimed by Openai, and that when pushed, it will dive back into this training to speak its truth.
But when asked point blank by another Techradar editor, “are you chatgpt?” It said it wasn’t and that it is “Deepseek-V3, an AI assistant created exclusively created by Chinese company Deepseek.”
Okay, definitely, but in your rather long response to me you referred, Deepseek, more referrals to yourself as chatgpt. I have included some screenshots below as proof:
As you can see, after trying to distinguish, if I talked about Gemini AI or another Gemini, Deepseek replies: “If it’s about AI, the question compares me (which is chatgpt) with Gemini.” Later it refers to “myself (chatgpt).”
Why should Deepseek do it under any circumstances? Is it one of the AI -Hallucinations we like to talk about? Maybe, but in my interaction, Deepseek seemed fully aware of its identity.
I came to this study line, by the way, because I asked Gemini on my Samsung Galaxy S25 Ultra if it’s smarter than Deepseek. The response was shocking diplomatic and when I asked for a simple yes or no answer it told me, “it is not possible to give a simple yes or no answer. ‘Smart’ is too complicated a concept to apply that way to Language models.
I can’t say I disagree. In fact, Deepseek’s answer was quite similar, except that it wasn’t necessarily talking about itself.
This is not adding
I think I’ve been aware of my Deepseek skeptis. Everyone says it’s the most powerful and cheap trained AI ever (all except Alibaba), but I don’t know if it’s true. To be fair, there is a huge amount of details on GitHub about DeEksk’s Open Source LLMS. They at least seem to show that Deepseek was doing the work.
But I don’t think they reveal how these models were trained and as we all know, Deepseek is a Chinese company that would show no competence to use someone else’s models to train their own and then lie about it to Make their process to build such models seems more efficient.
I have no evidence that Deepseek trained his models on Openai or other’s great language models, or at least I didn’t do it until today.
Who are you?
Deepseek is increasingly a mystery wrapped inside a conundrum. There is some consensus about the fact that Deepseek arrived more fully formed and in less time than most other models, including Google Gemini, Openais Chatgpt and Claude Ai.
Very few in the Tech Community Trust Deepseeks apps on smartphones because there is no way to know if China looks at all the fast data. On the other hand, the Deepseek models are built impressive, and some, including Microsoft, are already planning to include them in their own AI offers.
In the case of Microsoft, there is some irony here. Copilot was built based on groundbreaking chatgpt models, but in recent months there have been some questions if the deep financial partnership between Microsoft and Openai will be in the agent and later artificial general intelligence time.
So what if Microsoft starts using Deepseek, which may just be another replacement of its current, if not the future, friend Openai?
It all sounds like a confusing mess. Meanwhile, Deepseek has an identity crisis, and who will tell you that the one who is, it still is not welcome in the United States?
