Ex-Intel CEO Pat Gelsinger and also the current chairman of startup Gloo has eventually become a vocal advocate for DeepSeek R1 which is an innovative Chinese open source AI model that is challenging the dominance of OpenAI, reported Tom’s Hardware.
Budget with ET
Middle class, women, poor dominate Modi's pre-Budget speech. Relief ahead?
Economic Survey offers hope to 'aam aadmi' on inflation pains, subject to terms and conditions
India Inc earns fat profits but is not willing to pay more: Economic Survey
According to Tom’s Hardware, following the comparison of DeepSeek R1 with OpenAI's o1 model, ex-Intel CEO Pat Gelsinger chose R1 for the new AI service of Gloo named ‘Kallm’ solely due to its superior performance and obviously cost effectiveness.
DeepSeek has achieved some remarkable technological milestones by training its R1 model with significantly lower resources while comparing it to its competitors. The company legitimately claimed to have spent only $5.6 million in total training expenditure which is approximately 11 times less computational power than the Llama 3 405b model of Meta, asserted Tom’s Hardware.
Adding onto that, now by utilizing the Huawei Ascend AI accelerators for inferencing, DeepSeek has reduced the hardware dependency and operational costs as well.
In spite of the huge enthusiasm shown by ex-Intel CEO Pat Gelsinger, the model is eventually facing severe challenges which include investigations by Microsoft and OpenAI regarding the potential unauthorized data usage from ChatGPT, noted Tom’s