The Wall Street Journal recently reported on the competition in creating services from technology within the LLM group, such as ChatGPT. Many companies are heavily investing in this area, but several services are unable to generate profits, despite collecting usage fees.
According to WSJ, the average cost of providing GitHub Copilot service is still at $20 per person per month, but they only charge $10 for the service. Some users are utilizing it so much that they incur costs of up to $80 per month. With 1.5 million users, GitHub Copilot is generating reasonable revenue for Microsoft.
LLM-based systems, like GPT-4, provide satisfactory answers. However, they come with relatively high service costs. New services that utilize LLM as a feature are also becoming more expensive, such as Microsoft 365 Copilot, which charges $30 per person per month.
The high cost of service usage has forced various services incorporating LLM features to find ways to reduce costs. For example, Zoom is attempting to simplify its own model, while Adobe restricts usage. Even Bing Chat may consider switching to Meta’s Llama 2. Additionally, there is a trend to optimize the operation of large-scale models.
TLDR: The Wall Street Journal reports on the fierce competition among LLM-based services, with companies investing heavily. However, many struggle to turn a profit despite collecting user fees. GitHub Copilot is an exception, with 1.5 million users generating significant revenue for Microsoft. LLM-based systems provide satisfactory results, but their service costs remain high. As a result, companies are seeking cost-saving measures and exploring alternative solutions offered by Meta’s Llama 2. There is also a focus on optimizing the operation of large-scale models.