Will probably edit it, as many people complained. The focus here is useful predictions, in the sense of predictions that have utilitarian value, but string theory doesn't make any predictions at all, which is another issue altogether.