Unveiling Chessnut ChessMind: The Reasoning Behind a Subscription-based Chess AI Training
What is the subscription service on Evo? Which features require payment and which ones are free? What are the prices?
Many users have learned about our upcoming subscription service, Chessnut ChessMind Premium, through Youtubers' videos alongside the release of Chessnut Evo. Yes, due to the highly server-dependent nature of our customized AI training service, we will be charging for this service through a subscription model.
The training service is deployed on Amazon Web Services, utilizing Amazon EC2 instances of g5.4xlarge, which are equipped with a 16-core CPU and NVIDIA A10G Tensor Core GPU. When users upload game data of chess players for customized chess AI engine training, the service utilizes the GPU for training. This process typically takes around 40 to 60 minutes, depending on the uploaded data. The cost for a single server is approximately $1,000 per month. Initially, we plan to enable 10 servers, and we will increase the number of servers as the number of subscriptions grows.
The subscription prices are set at $19.9 per month and $99.9 per year, allowing subscribers to enjoy unlimited training sessions throughout their subscription period.
However, recognizing that not all users may require such a high frequency of training, we have listened to user feedback and introduced a pay-per-use option. Users can now choose to pay $4.9 per session, with an added bonus of two complimentary sessions for every purchase of a 10-session package. Additionally, customers who have purchased Chessnut Evo will receive ten free trial sessions.
Apart from the training service that relies on server computational power, all other features on Chessnut Evo are provided at no additional cost. This includes the usage of Chess Engines such as Stockfish, Maia, and other engines, as well as the utilization of Chessnut Vision's image recognition technology. Furthermore, any customized engines created through the training service can be used for free even after the subscription expires.
We understand the value and significance of delivering a high-quality and customizable AI training service to our users. By implementing a subscription model, we can ensure the continuous improvement and expansion of our service, as well as cover the costs associated with the deployment and maintenance of powerful servers like the Amazon EC2 instances with NVIDIA A10G Tensor Core GPUs.
Chessnut ChessMind Premium aims to provide chess enthusiasts with an exceptional and tailored training experience, empowering them to enhance their skills and achieve greater success in the game. We believe that our subscription pricing structure offers a fair and reasonable approach, catering to different user needs while sustaining the development and advancement of our service. We are excited to have you join us on this chess journey and look forward to helping you unlock your full chess potential with Chessnut ChessMind.
Furthermore, we are committed to continuously improving and expanding our services in the future. As we grow, we plan to introduce additional features and services related to Chess Engines and AI, without imposing any extra charges on our subscribed users. Our goal is to provide ongoing value and enhance your chess experience as a valued member of the Chessnut ChessMind community.
4 comments
For the AI training service, will it have the option to aim for a combination of all the opponents I have played to create something close to the overall style of play I have encountered online instead of just one specific player?
As for the sustainability mentioned in the other post here, I agree that AWS services should be scaled more based on expected users vs what a given training session requires. There are ways to set up autoscaling with EC2 so you can potentially save money by not constantly running as many server instances if the user load doesn’t demand all of the planned resources 24/7. Using other instance options outside of entirely on-demand or maybe using Lambda functions or EKS/ECS containers might also be a more viable approach depending on the use cases on the backend and the work required to port those over from a more traditional server setup.
I have no idea what the premium service provides. Links to more information about details of the service would have been useful. Thanks.
Concerning the subscription-based AI Bot creation: I would like to suggest that after your current campaign ends, you survey users of all three of your products (Chessnut Air, Chessnut Pro, and Chessnut Evo) and ask them to select 20 players, past and present, that they would like to see AI Bots created from their games. On the list you might include all of the past World Chess Champions, other top Grandmasters, popular IM Chess Trainers. After identifying the top 10-20 choices, announce and create AI Bots for them and make them available for sale to the users so that you can have conservation of server use so that for example you don’t have 100 people all creating AI Bots of Magnus Carlsen’s games, but you create an AI Bot of Magnus Carlsen’s games and then make it available to those who want it. I would also like to suggest that after you create AI Bots of some of the world’s top players, that you create a tournament playing the Bots against each other. I know I would greatly appreciate it and I think other users would appreciate conserving server use and not duplicating our efforts. I think this would also greatly increase your sales of the AI Bot service, especially for some people who may not know what they want and may be reluctant to begin using the service.
Thanks for making it clear.
I cannot believe your plan is sustainable. You say you will pay $10,000 USD per month, but only sold 500 Evo boards. Your math is not good.
Nice to know you’re trying.