Saturday AI can make mistakes

In trying to be a helpful assistant, Saturday AI can occasionally give answers that are incorrect or misleading.

This is known as “hallucinating,” and it comes from the current limits of generative AI models, including the ones Saturday AI uses. For example, a product’s formula or label may have changed since Saturday AI learned about it, so it can describe the product as it used to be. When it reads a photo of a nutrition label, it can misread a number or a unit. It can also give a number, such as grams of carbohydrate or milligrams of sodium, that sounds certain and is wrong. In other words, Saturday AI can write things that look correct but are mistaken.

Don’t rely on Saturday AI as your only source of truth, and look carefully at any advice that matters most, such as advice on fueling in the heat or for an effort that lasts many hours. If you have a medical condition or take medication, check its advice with a healthcare professional.

When an answer uses activities synced from a device, it can show a Sources line naming where that data came from, such as a Garmin or COROS watch. Check the activity itself in Saturday: it can hold context or details the answer leaves out. The quality of Saturday AI’s answers also depends on the data it reads, so checking the original helps you catch anything it misread or took out of context. When it reads a nutrition label for you, compare its numbers with the package before you save the product.

You can tap the thumbs down under a reply to let us know it was unhelpful, or write to us at support@saturdaymorning.fit with your thoughts or suggestions.

To learn more about what Saturday AI shares and when, read our app privacy policy.