‘Pokémon Go’ players unknowingly trained delivery robots with 30 billion images

*Pokémon Go Players Unknowingly Contribute to AI Training*

A recent report has revealed that millions of Pokémon Go players have inadvertently contributed to the training of delivery robots through the use of their mobile devices. According to the report, players of the popular augmented reality game have captured and uploaded over 30 billion images of their surroundings to the game's servers.

*The Connection to AI Training*

While the primary purpose of these images was to aid in the game's augmented reality features, they have also been used to train artificial intelligence (AI) models. One of the companies that has been using these images is Nuro, a robotics company that has developed autonomous delivery robots. These robots use computer vision and machine learning algorithms to navigate and deliver packages.

*How the Images Were Used*

Nuro's AI models were trained on a dataset of images collected from various sources, including Pokémon Go players. The images were used to teach the AI to recognize and identify objects in the environment, such as roads, sidewalks, and buildings. This training data enabled the AI to improve its accuracy and reduce errors in navigation.

*Implications and Concerns*

The use of Pokémon Go images to train AI models raises several implications and concerns. Firstly, it highlights the unintended consequences of playing popular mobile games. Players may not be aware that their actions are contributing to the development of AI models that may have far-reaching applications.

Secondly, the use of user-generated content to train AI models raises questions about data ownership and consent. While players may have agreed to the game's terms of service, they may not have explicitly consented to their images being used for AI training.

Lastly, the reliance on user-generated content to train AI models may also lead to biases and inaccuracies in the AI's decision-making processes. If the images used to train the AI are representative of a particular demographic or geographic region, the AI may learn to recognize and respond to features that are unique to that area, potentially leading to errors in other contexts.

*Conclusion*

The use of Pokémon Go images to train AI models is a fascinating example of the unintended consequences of playing popular mobile games. While the game's developers and the companies that use the images may have intended for the images to be used for AI training, the implications and concerns raised by this practice highlight the need for greater transparency and accountability in the development and use of AI models.