function vs property in interface

• Declare as a property whose type is function

interface Logger {
    log: (message: string) => void;
}

• Declare as a normal function

interface Logger {
    log(message: string): void;
}

Differences

1. If the method is declared as a interface function, then it's possible for you to add more overload versions.

   interface Logger {
       log(message: string): void;
   }
   
   // In other places
   interface Logger {
       log(message: string, level: string): void;
   }

   Declaring method as a property, on the other hand, prevents you from duplicating the property declarations which have different types:

   interface Logger {
       log: (message: string) => void;
   }
   
   // Does not work
   interface Logger {
       log: (message: string, level: string) => void;
   }

2. The readonly modifier only has effect with the property declaration.

   interface Person {
       firstName: string;
       lastName: string;
   
       readonly fullName: () => string;
   
       // Doesn't work
       // readonly fullName(): string;
   }

3. TypeScript generates different output for a class that implements the interface methods.
   Assume that we have a class ConsoleLogger that simply logs the message in the Console window.
   For the first approach:

   interface Logger {
       log: (message: string) => void;
   }
   
   class ConsoleLogger implements Logger {
       log = (message: string) => {
           console.log(message);
       };
   }
   
   // Generated JavaScript code:
   //
   // class ConsoleLogger {
   //    constructor() {
   //        this.log = (message) => {
   //            console.log(message);
   //        };
   //    }
   // }

   For the second approach:

   interface Logger {
       log(message: string): void;
   }
   
   class ConsoleLogger implements Logger {
       log(message: string) {
           console.log(message);
       }
   }
   
   // Generated JavaScript code:
   //
   // class ConsoleLogger {
   //    log(message) {
   //        console.log(message);
   //    }
   // }

   Looking at the generated JavaScript codes, you'll see the different outputs.
   The first approach produces a property log in the constructor. It means that log will be created every time you create a new instance of class.
   While the second approach produces the log method, and it exists in all instances of class. The log method also is a member of class prototype, so we can extend the class to override the method if needed:

   class ConsoleLogger implements Logger {
       log(message: string) {
           console.log(message);
       }
   }
   
   class ConsoleLoggerWithColor extends ConsoleLogger {
       // Override the `log` method
       log(message: string) {
           // Display the message in white color and blue background area
           console.log('%c%s', 'color: white; background: blue', message);
       }
   }

   See the differences between declaring methods in class constructor and prototype for more details.