Apple downloads ~45 TB of models per day from our S3 bucket (twitter.com)

Article summary

The creator of a machine learning model reported that Apple is downloading approximately 45 TB of models per day from their S3 bucket, resulting in significant bandwidth costs. The creator jokingly asked if Apple is trying to bankrupt them. The models are publicly available, but the high volume of downloads is costly. The situation highlights the issue of bandwidth costs for publicly available data.

Main themes

• bandwidth costs
• publicly available data
• machine learning models
• cloud storage
• data caching

What commenters say

• The company should not complain about the costs if they made the data publicly available without taking steps to prevent or charge for excessive downloads.
• The company can use services like Cloudflare to reduce bandwidth costs and potentially pass the costs on to the requester.
• Apple's high volume of downloads may be due to a misconfigured build pipeline or server farm, rather than individual employees downloading the models.
• The company should consider contacting Apple to discuss potential payment or caching solutions rather than publicly complaining about the issue.
• Using a content delivery network (CDN) like Cloudflare can help reduce bandwidth costs, but may not be suitable for very large volumes of data.
• Some commenters believe that the company is using the situation as a publicity stunt to attract attention to their product.
• Others argue that the company's handling of the situation may harm their business prospects and reputation.