April 27, 2023
According to a recent report by SemiAnalysis, OpenAI is spending an estimated $694,444 per day to keep its Chatbot ‘ChatGPT’ operational.
The system is thought to consist of 3,617 HGX A100 servers with 28,936 GPUs at a cost of roughly 0.36 cents per query. However, Dylan Patel, Chief Analyst at SemiAnalysis, cautions that expenses for the new GPT-4 model, which is projected to be more expensive to operate than its predecessor, might be much higher.
The power consumption of the specialized processors necessary to run the system is one of the key reasons for the high pricing. As a result, Microsoft, one of OpenAI’s top stockholders, is rumored to be building its own AI processor called ‘Athena,’ which might replace NVIDIA GPUs and dramatically lower operating costs.
The investigation revealed that the code generated by ChatGPT is not safe enough for most cases, and the chatbot itself admitted as much when queried about the code’s security. The researchers requested ChatGPT to create 21 programs and scripts in various programming languages such as C, C++, Python, and Java in order to assess its coding ability.
The AI chatbot initially only managed to build five secure programs, but with extra assistance from the researchers, it was able to construct an additional seven secure code snippets. Although the chatbot’s coding ability is concerning, greater research may assist establish its future potential for safe and productive application.
The sources for this piece include an article in TechSpot.
