What is the point of a nil AutoreleasingUnsafeMutablePointer in a Swift closure?

I was reading the different ways to parse REST API calls in Swift and came across the following:

var url : String = "http://ift.tt/1tENtnZ"
var request : NSMutableURLRequest = NSMutableURLRequest()
request.URL = NSURL(string: url)
request.HTTPMethod = "GET"

NSURLConnection.sendAsynchronousRequest(request, queue: NSOperationQueue(), completionHandler:{ (response:NSURLResponse!, data: NSData!, error: NSError!) -> Void in
    var error: AutoreleasingUnsafeMutablePointer<NSError?> = nil
    let jsonResult: NSDictionary! = NSJSONSerialization.JSONObjectWithData(data, options:NSJSONReadingOptions.MutableContainers, error: error) as? NSDictionary

    if (jsonResult != nil) {
        // process jsonResult
    } else {
       // couldn't load JSON, look at error
    }


})

The one line that makes no sense to me is var error: AutoreleasingUnsafeMutablePointer<NSError?> = nil. We already captured our NSError parameter stored in a variable called error, and now we're overwriting that and making it nil in our first line in the closure? Or if somehow Swift then performs a downcast of the error from type NSError! to AutoreleasingUnsafeMutablePointer<NSError?>, then can someone explain how that happens?

Thanks!