Chinese language tech firm DeepSeek is as soon as once more underneath scrutiny, this time over allegations that it could have used Google’s Gemini AI to coach its newest mannequin. Beforehand, the corporate confronted criticism for allegedly utilizing ChatGPT outputs to coach its earlier variations.
Lately, DeepSeek launched an up to date model of its AI mannequin, R1, praised for its spectacular reasoning skills. Nonetheless, questions have arisen relating to the dataset used to coach the mannequin. In line with Australian developer Sam Paech, the mannequin’s outputs bear an uncanny resemblance to these produced by Gemini 2.5 Professional, together with comparable phrase selections and phrasing. Whereas Paech admits this isn’t definitive proof, he emphasizes the putting similarities.
A Gemini-like Reasoning Model
One other developer identified that DeepSeek’s mannequin shares a really comparable “reasoning type” to Gemini. This statement is predicated on the AI’s intermediate steps when fixing issues—suggesting a sample that carefully mirrors how Gemini approaches reasoning duties.
This isn’t the primary time DeepSeek has confronted such allegations. In an earlier controversy, builders famous that DeepSeek’s V3 mannequin would often discuss with itself as “ChatGPT,” elevating issues that it could have been educated on OpenAI’s proprietary dialogue knowledge.
Whereas coaching fashions utilizing the outputs of extra superior AIs isn’t a brand new tactic within the AI group, most corporations—particularly OpenAI—explicitly prohibit such practices. Nonetheless, implementing these restrictions seems to be more and more tough.
Comply with us on TWITTER (X) and be immediately knowledgeable in regards to the newest developments…
Copy URL
Comply with Us
