Serving technology enthusiasts for more than 25 years. TechSpot is a trusted source for tech advice and analysis.
A hot topic: Microsoft raises eyebrows by announcing it will host DeepSeek R1 in its Azure cloud service. Microsoft is investigating the allegations that DeepSeek violated its terms of service after using ChatGPT outputs for training its system.
DeepSeek R1 made waves in the AI community when it was launched last week. DeepSeek, a Chinese developer, touted the model as a free simulated reasoning model with performance comparable to OpenAI’s O1 but at a fractional cost. OpenAI’s o1 model costs $60 per million tokens of output, while DeepSeek lists its R1 for just $2.19 – a striking contrast that has sunk the stock prices of AI-related companies like Nvidia. Microsoft’s decision,
to host R1 on Azure does not seem too outlandish. The tech giant has already Azure AI Foundry offers more than 1,800 AI models, giving developers access for experimentation and integration. Microsoft does not discriminate, since it benefits from any AI platform that runs on its cloud infrastructure. The decision is ironic, since OpenAI (a Microsoft-backed company with which Microsoft has partnered) has spent the past week criticizing the model used to distill ChatGPT outputs.
DeepSeek R1 is here.Performance comparable to OpenAI-o1 model
Fully Open-Source Model & Technical Report
MIT license: Distill & Commercialize freely!The website and API are now live! DeepThink is now available at https://t.co/v1TFy7LHNy today!
– Deepseek (@deepsek_ai)””https://twitter.com/deepseek_ai/status/1881318130334814301?ref_src=twsrc%5Etfw””> OpenAI claims the AI startup on January 20, 2025 (19659009) Fox News reported that violated its terms of service when it used “distillation,” . Distillation occurs when developers train an AI using outputs of a more advanced AI system. Users began to have suspicions after It was discovered an earlier model of DeepSeek V3 sometimes referred to themselves as “ChatGPT,” indicating that DeepSeek used OpenAI generated data to fine-tune their system.
This move seems a bit hypocritical considering that Microsoft security researchers launched an ethics investigation into DeepSeek on Wednesday. Anonymous sources claim the investigation is focused on whether DeepSeek extracted significant amounts of data using OpenAI’s API in the fall of 2024.
OpenAI CEO Sam Altman publicly welcomed DeepSeek despite his frustrations. Altman acknowledged the cost-effectiveness of R1 in a Monday tweet, calling it “an impressive model” and promising that OpenAI would deliver soon “much better results.” According to analysts, the company could release a new product, o3 mini, as early today.
The r1 from deepseek is a very impressive model. Especially in terms of what they can deliver for the price.
We will deliver much better models, and it’s also legitimatingly invigorating to be a new rival! We will remove some releases.
Sam Altman (@sama), January 28, 2025.
OpenAI’s outcry against DeepSeek’s data practices is noteworthy given its own history with alleged data abuse. The New York Times filed a lawsuit against OpenAI, Microsoft and others for using copyrighted journalistic content without permission. OpenAI has also made deals with online communities and publishers, such as The associated Press to access user-generated information for training.
This whole situation exposes AI industry’s hypocritical attitude towards data ownership. Andreessen Horowitz Investments, another Open AI investor argued in a legal filing for 2023 that training AI models shouldn’t be considered copyright violation, as they merely “extract information” existing works. If OpenAI really believes in this principle, then DeepSeek simply follows the same rules.
Currently, the AI industry is a more or less free-for-all. We don’t have any laws to regulate AI directly. And those that do, like copyright laws and trade laws, can be interpreted in a way that is favorable by the AI companies that break them.
