OpenAI has accused the Chinese AI startup DeepSeek of improperly using its data to develop the R1 model, a rival AI model. The accusation centers on DeepSeek's alleged use of a technique called 'distillation,' which involves training smaller AI models using the outputs of larger, more sophisticated models like those developed by OpenAI. This practice is claimed to have been used to train DeepSeek's R1 model, which has been noted for its efficiency and performance despite requiring fewer resources than its competitors. Concurrently, DeepSeek has reported facing significant cyberattacks since January 3, with Chinese state media attributing the origin of these attacks to US IPs. These cyberattacks, which intensified recently, included DDoS attacks and brute-force attacks that have disrupted DeepSeek's services, prompting the company to temporarily limit new user registrations. The situation has sparked a debate on intellectual property rights and the ethics of model distillation in the AI industry, with Microsoft also involved in probing the potential data breach linked to DeepSeek and the use of the ChatGPT data set.
It’s completely fucked up for DeepSeek to train using stolen output from OpenAI after all the time OpenAI invested in stealing copyrighted content and Elon’s initial donation to create that model
Chinese state-linked accounts hyped DeepSeek AI launch ahead of US stock rout, Graphika says - https://t.co/1PQf1j7ZTC
AI that can identify ChatGPT-generated text suggests DeepSeek might be a copy https://t.co/uo605R73Jj