Upsizing merely increases the pixel count without any increase in detail. IOW, by upsizing pixel count by 10X on a print which is 10X larger, you merely lower the eye's ability to see individual areas of color.
Think of a photo in a newspaper vs. the same photo on the pages of National Geographic...you easily see 'dots' in the newspaper due to the 'screen' putting dots of ink on the page for color, whereas it is more difficult to see 'dots' on National Geographic. In neither case, though, might you be able to see individual eyebrow hairs.
Stated differently, in terms of detail resolution seen in a photo, if you take a top quality lens listed by photozone.de as having MTF of 3400 line-pairs of resolution in a 22MPixel camera (5DIII), increasing its vertical pixel count from native 3840 pixels to 7680 pixels still leaves only 3400 line-pairs seen in the photo. The line edges of a line running diagonally across the photo might look less like a staircase and more like a line, though.
Using bicubic interpolation would be a perfectly fine way to get a pixel count increase, to reduce the stairstep appearance, but the edge appears less sharp, as illustrated in this example

Here is a 200x200 pixel horizontal line with arrowhead, and a diagonal line...
And here is an upres of the above image to 400x400 ...