Earlier this month, Figma was forced to suspend its AI assistant for designers, Make Designs. This happened after the assistant began creating projects for users that were very similar to Apple applications for the iOS platform. Now the company’s specialists have studied this issue and explained why the neural network created such projects.
Image source: Figma
Using app designs similar to Apple products could lead to legal troubles, so users of the Figma platform expressed dissatisfaction with the performance of the software solution. In addition, the incident suggests that the Figma neural network could be trained on designs from Apple and other companies. It turned out that the problem arose because Figma engineers did not properly test the components being added to the generative neural network.
Figma’s announcement notes that the company “closely studied” the underlying Make Designs systems during algorithm development and beta testing. However, a week before the launch of the neural network into public access, new components and example layouts were added to the configuration, which were not properly tested. The company admits that some of these components were similar to real-life applications, and they appeared as a result of the AI algorithm processing certain user requests.
As soon as Figma became aware of the problem with the design algorithms, the components that caused the generated content to resemble real-life applications were removed. The company is currently working to improve quality control, once completed, Make Designs will once again be publicly available. No specific time frame was announced for when the AI assistant will be publicly available again.