G'day Darmin,
Out of the box you'd need to surface the S3 content via a fileystem of some kind (e.g. EBS or EFS), so that the tool is able to read it via standard Java (POSIX) file I/O. The tool would then physically copy the content into whatever storage system you've configured Alfresco to use.
Alternatively, you could implement a custom
BulkImportSource that would read S3 directly (this could also support "in place" imports, if your Alfresco contentstore is also in S3). Symflofy (who embed the bulk import tool in their software) have implemented this, but I don't believe they've open sourced it (there've been some recent email threads here discussing this).
Cheers,
Peter